[ { "name": "1Password AI Agent Identity Kit", "id": "1password-ai-agent-identity-kit", "companyName": "1Password", "companyUrl": "https://1password.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Security engineer", "Platform engineer", "IT security manager" ], "problemsSolved": [ "shared API keys used by agents", "scoping agent access to approved credentials", "auditing agent credential use" ], "sources": [ { "id": "s1", "url": "https://1password.com/solutions/ai-agent-identity", "title": "AI Agent Identity Kit - 1Password", "publisher": "1Password", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-04-29", "supports": "Describes verifiable agent identity, short-lived agent tokens, scoped authorization and audit trail linkage, layered on existing vaults and SDK." }, { "id": "s2", "url": "https://1password.com/press/2026/mar/1password-unified-access", "title": "1Password Launches Unified Access for AI Agent Security", "publisher": "1Password", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-03-17", "supports": "States Unified Access Pro is generally available covering human, machine and AI agent identities, with Audit coming soon." } ], "claims": [ { "id": "c1", "statement": "The kit assigns a verifiable identity to a software agent rather than only the human or service account behind it, tied to a specific workload identity with short-lived credentials.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Solution page; no reference documentation on attestation or supported runtimes was fetched.", "capability": "verifiable-agent-identity" }, { "id": "c2", "statement": "1Password Unified Access Pro became generally available on March 17, 2026 with Discover and Secure capabilities available at launch; Audit was described as coming soon and runtime scoped credentials for agent and machine workloads as a later expansion.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Vendor press release; roadmap items are announced rather than shipped.", "capability": "unified-access-availability" } ], "limitations": [ "Product naming is inconsistent across vendor pages (AI Agent Identity Kit vs Agent Identity Toolkit)." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "CA" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "issue": "Vendor page uses 'AI Agent Identity Kit' and 'Agent Identity Toolkit' interchangeably.", "sourceIds": [ "s1" ] } ], "editorialNotes": "Recorded the agent-specific kit rather than the whole 1Password Extended Access Management suite." }, { "name": "Aembit Workload IAM", "id": "aembit-workload-iam", "companyName": "Aembit", "companyUrl": "https://aembit.io", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Platform engineer", "Security architect", "DevSecOps lead" ], "problemsSolved": [ "removing static secrets from agents and workloads", "conditional access for workload-to-workload calls", "centralized access logging for non-human traffic" ], "sources": [ { "id": "s1", "url": "https://aembit.io/product/", "title": "Product Overview | Aembit", "publisher": "Aembit", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Aembit Workload IAM: secretless workload authentication via attestation, global policy system, credential providers and unified access logs." }, { "id": "s2", "url": "https://aembit.io/blog/securing-ai-agents-without-secrets/", "title": "Workload Identity for AI Agents: Securing AI Agents Without Static Credentials", "publisher": "Aembit", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2025-09-09", "supports": "Describes credential injection, OAuth client-credentials issuance and context-aware policy evaluation specifically for AI agents." }, { "id": "s3", "url": "https://docs.aembit.io/get-started/use-cases/ai-agents/", "title": "Securing AI agent access to your resources", "publisher": "Aembit", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Official Aembit product documentation stating that AI agents are treated as workloads, and that Aembit's attestation methods, SPIFFE compatibility, credential issuance and access policies extend to agentic AI." } ], "claims": [ { "id": "c1", "statement": "Aembit uses client environment attestation for secretless workload authentication so the client does not need a long-lived identity secret.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page does not enumerate attestation methods per platform.", "capability": "secretless-workload-auth" }, { "id": "c2", "statement": "For AI agents, a credential-injection layer sits between the agent and the API endpoint and issues short-lived tokens (described as typically expiring in 15-30 minutes) after policy evaluation.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Blog post, not reference documentation; token lifetimes described as typical rather than configurable guarantees.", "capability": "credential-injection-for-agents" }, { "id": "c3", "statement": "A global policy system defines access between non-human identities across environments and logs access attempts in a single format viewable in Aembit or a SIEM.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documented retention period or log schema.", "capability": "access-policy-and-logging" }, { "id": "c4", "statement": "Aembit documents AI agents as workload identities and applies its attestation, credential issuance and access policy mechanisms to them.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "The documentation describes the mechanism and its applicability; it does not establish coverage of any specific agent framework, nor measured enforcement outcomes.", "capability": "agent-as-workload-identity" } ], "limitations": [ "No evidence of agent discovery, inventory or entitlement review features." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Agent-specific evidence comes from a vendor blog; core product page speaks to non-human workloads generally. Strengthened during merge with the official docs.aembit.io AI-agents use-case page, which is also the source cited by DutyGraph's existing live directory entry." }, { "name": "agentgateway", "id": "agentgateway", "companyName": "Solo.io (project hosted by the Linux Foundation)", "companyUrl": "https://www.solo.io/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "observability" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Platform engineering lead", "CISO", "API platform architect" ], "problemsSolved": [ "single ingress for MCP and LLM traffic", "per-tool access control", "agent traffic audit" ], "sources": [ { "id": "s1", "url": "https://agentgateway.dev/", "title": "agentgateway | Agent Connectivity Solved", "publisher": "Linux Foundation / agentgateway project", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes agentgateway as an open source HTTP and gRPC gateway covering service, LLM, MCP and A2A traffic, donated by Solo.io and hosted by the Linux Foundation." }, { "id": "s2", "url": "https://docs.solo.io/agentgateway/2.3.x/quickstart/mcp/", "title": "MCP servers | Solo.io documentation", "publisher": "Solo.io", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents proxying an MCP server via agentgateway on Kubernetes with appProtocol agentgateway.dev/mcp, mcp-path annotation, AgentgatewayBackend and HTTPRoute, and verification with MCP Inspector." }, { "id": "s3", "url": "https://www.solo.io/products/agentgateway", "title": "Agentgateway: The AI-Native Gateway", "publisher": "Solo.io", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists native MCP OAuth 2.1, tool-level RBAC, secure token exchange, cryptographic audit trails, rate limits and budget enforcement." } ], "claims": [ { "id": "c1", "statement": "agentgateway is an open-source HTTP/gRPC gateway that handles service traffic, LLM provider traffic, MCP tool traffic and agent-to-agent communication in one data plane.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The project page does not quantify enforcement latency or list supported policy syntaxes.", "capability": "unified-agent-data-plane" }, { "id": "c2", "statement": "Kubernetes-based configuration is documented for proxying an MCP server through agentgateway, including an MCP-specific appProtocol, MCP path annotations and an AgentgatewayBackend resource.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The MCP quickstart documents connectivity and tool invocation only; it does not document authorization policy, guardrails or human approval.", "capability": "mcp-proxy-config" }, { "id": "c3", "statement": "The vendor product page states agentgateway provides native MCP OAuth 2.1, tool-level RBAC, secure token exchange and cryptographic audit trails.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "These controls are asserted on a marketing product page rather than in reference documentation fetched here.", "capability": "tool-level-rbac" } ], "limitations": [ "Authorization policy detail is thin in the fetched documentation", "Enterprise control plane features are packaged separately as Solo Enterprise for agentgateway" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "hybrid" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Single entry covering both the community project (agentgateway.dev, Linux Foundation hosted) and Solo.io's commercial packaging; kept as one product because the enforcement surface is the same gateway. Also surfaced in the open-source pass; runtime-controls kept as primary." }, { "name": "AgentOps", "id": "agentops", "companyName": "AgentOps", "companyUrl": "https://www.agentops.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineer", "Agent developer" ], "problemsSolved": [ "session-level agent monitoring", "tool-call and error visualisation", "LLM spend tracking", "debugging framework-based agents" ], "sources": [ { "id": "s1", "url": "https://docs.agentops.ai/v2/introduction", "title": "AgentOps: Introduction", "publisher": "AgentOps", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes observability and monitoring for AI agents and LLM apps, sessions per execution, session waterfall with LLM calls, action events, tool calls and errors, prompt/completion inspection, event breakdowns and timings, session drawer and overview, and custom trace decorators." }, { "id": "s2", "url": "https://github.com/AgentOps-AI/agentops", "title": "AgentOps-AI/agentops", "publisher": "GitHub / AgentOps", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States MIT license for the AgentOps app, self-hosting of dashboard plus API backend on the customer's cloud, analytics on LLM calls, spend tracking with foundation model providers, and framework integrations including OpenAI Agents SDK, CrewAI and AG2/AutoGen." } ], "claims": [ { "id": "c1", "statement": "Each execution of the program is recorded as a session, with a Session Waterfall visualising LLM calls, action events, tool calls and errors, drilldowns showing total execution time and exact prompt and completion for a given LLM call, plus event-type breakdowns and custom traces via a decorator.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Multi-agent handoffs, datasets, human feedback and evaluations are not documented on the introduction page.", "capability": "session-recording" }, { "id": "c2", "statement": "The repository states the AgentOps app is open source under the MIT license and that the full app including dashboard and API backend can be run on the customer's own cloud, with integrations listed for the OpenAI Agents SDK, CrewAI and AG2/AutoGen and spend tracking for foundation-model providers.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "OpenTelemetry support, RBAC, SSO, audit logs, retention and compliance are not stated; the latest listed release predates 2026.", "capability": "license-and-self-hosting" } ], "limitations": [ "No documented access control, audit trail, retention configuration or redaction: unsuitable on current evidence as a governance record system.", "Release cadence evidence on the repository predates 2026, so current maintenance level is unclear." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "CrewAI", "type": "documented_api", "sourceIds": [ "s2" ], "limitations": "Listed as a supported framework integration; instrumentation depth not documented." } ], "conflicts": [], "editorialNotes": "Hybrid: MIT-licensed self-hostable app plus hosted dashboard." }, { "name": "AGNTCY Identity", "id": "agntcy-identity", "companyName": "AGNTCY", "companyUrl": "https://agntcy.org", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "IAM manager", "Platform architect", "CISO" ], "problemsSolved": [ "unique identity for agents and MCP servers", "verifiable agent credentials", "cross-vendor agent trust" ], "sources": [ { "id": "s1", "url": "https://github.com/agntcy/identity", "title": "agntcy/identity", "publisher": "AGNTCY / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes identifier issuance, verifiable credentials, badge creation/verification, external identity onboarding and key storage options." }, { "id": "s2", "url": "https://www.linuxfoundation.org/press/linux-foundation-welcomes-the-agntcy-project-to-standardize-open-multi-agent-system-infrastructure-and-break-down-ai-agent-silos", "title": "Linux Foundation welcomes the AGNTCY project", "publisher": "Linux Foundation", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms Linux Foundation hosting and the identity/discovery/messaging/observability scope." }, { "id": "s3", "url": "https://api.github.com/repos/agntcy/identity", "title": "GitHub REST API repository record", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Reports Apache-2.0 license, archived=false, last push 2026-02-24." } ], "claims": [ { "id": "c1", "statement": "The repository states AGNTCY Identity assigns universally unique identifiers backed by verifiable credentials to agents, MCP servers and multi-agent systems, supports bringing identities from providers such as Okta, A2A agent cards or W3C DIDs, and can create, publish and verify badges.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "README-level documentation; no production deployment references, scale data or security review evidenced.", "capability": "identity-issuance-and-verification" }, { "id": "c2", "statement": "The Linux Foundation announced it welcomed the AGNTCY project, describing open source infrastructure for discovery, identity, messaging and observability across multi-agent systems and interoperability with A2A and MCP.", "support": "independently_supported", "sourceIds": [ "s2" ], "limitations": "Foundation hosting covers the umbrella project; it does not certify the identity component's security or maturity.", "capability": "foundation-hosting" }, { "id": "c3", "statement": "GitHub metadata records Apache-2.0 licensing, a non-archived repository, and last push February 2026.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Slower recent commit activity than other projects reviewed; roadmap and release status unclear.", "capability": "license-and-activity" } ], "limitations": [ "No evidenced enterprise deployments", "Interop with enterprise IdPs documented only at README level" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Cisco originated AGNTCY before Linux Foundation hosting; parentCompany left null because the project is foundation-hosted." }, { "name": "AI Incident Database (AIID)", "id": "ai-incident-database", "companyName": "Responsible AI Collaborative", "companyUrl": "https://incidentdatabase.ai", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "evaluation" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI risk manager", "Policy analyst", "AI safety researcher" ], "problemsSolved": [ "AI harm precedent research", "risk taxonomy inputs", "incident reporting culture" ], "sources": [ { "id": "s1", "url": "https://incidentdatabase.ai/", "title": "Artificial Intelligence Incident Database", "publisher": "Responsible AI Collaborative", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the database's purpose, submission process and incident listings." }, { "id": "s2", "url": "https://github.com/responsible-ai-collaborative/aiid", "title": "responsible-ai-collaborative/aiid", "publisher": "Responsible AI Collaborative / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the application architecture, public GraphQL endpoint and taxonomy support." } ], "claims": [ { "id": "c1", "statement": "The database states it indexes the collective history of harms or near harms realised in the real world by deployment of AI systems, invites public incident report submissions, and makes submissions indexed and discoverable.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Entries are derived from public reporting and community submission, so absence of an incident does not mean none occurred.", "capability": "incident-catalogue" }, { "id": "c2", "statement": "The repository states the application provides a read-only public GraphQL endpoint at /api/graphql, syncs content to an Algolia search index, and supports taxonomies for organising database content.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "API stability, rate limits and data licensing terms are not stated on the pages reviewed.", "capability": "programmatic-access" } ], "limitations": [ "GitHub license metadata is NOASSERTION, so reuse terms need checking", "Coverage biased toward English-language media reporting" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Distinguished from unaffiliated look-alike repositories using the same name; the canonical project is responsible-ai-collaborative/aiid." }, { "name": "Akeyless SecretlessAI", "id": "akeyless-secretlessai", "companyName": "Akeyless", "companyUrl": "https://www.akeyless.io", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Security architect", "Platform engineer", "CISO" ], "problemsSolved": [ "credentials leaking into agent context or memory", "just-in-time access for agents", "auditing agent credential use" ], "sources": [ { "id": "s1", "url": "https://www.akeyless.io/secrets-management/secretless-ai-agents/", "title": "Implementing Secure AI Agents with Akeyless SecretlessAI", "publisher": "Akeyless", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-09-04", "supports": "Describes brokered, on-demand, policy-scoped access for agents with credentials never entering the agent, plus centralized audit." }, { "id": "s2", "url": "https://www.akeyless.io/press-release/akeyless-unveils-breakthrough-solution-to-secure-ai-agent-identities/", "title": "Akeyless Unveils Breakthrough Solution to Secure AI Agent Identities", "publisher": "Akeyless", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-10-29", "supports": "Announces AI Agent Identity Security including SecretlessAI, AI Agent Identity Provider, AI Agent PAM and AI Insights." } ], "claims": [ { "id": "c1", "statement": "SecretlessAI gives AI agents access to enterprise systems without exposing passwords, API keys, secrets, certificates or keys, brokering every connection through the Akeyless Gateway and keeping credentials out of the agent's context, code, environment and memory.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; no reference documentation on supported connectors or failure modes was fetched.", "capability": "secretless-agent-access" }, { "id": "c2", "statement": "Akeyless announced an AI Agent Identity Security solution including an AI Agent Identity Provider, secretless authentication, identity federation and just-in-time access for agents.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Press release from October 2025; availability stage of each component is not stated.", "capability": "agent-identity-provider-and-jit" }, { "id": "c3", "statement": "The product maintains a centralized audit trail of credential use and agent access across enterprise systems.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Log schema and retention are not documented on the fetched page.", "capability": "central-agent-audit-trail" } ], "limitations": [ "No independent reporting or documentation was fetched to corroborate the secretless architecture." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Entry scoped to the agent-specific SecretlessAI / AI Agent Identity Security offering, not the general secrets management platform." }, { "name": "Amazon Bedrock AgentCore", "id": "amazon-bedrock-agentcore", "companyName": "Amazon Web Services", "companyUrl": "https://aws.amazon.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "identity-access", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Cloud architect", "Platform engineering lead", "AI engineering manager" ], "problemsSolved": [ "running agents securely at scale", "turning APIs into agent tools", "agent identity and credential management" ], "sources": [ { "id": "s1", "url": "https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html", "title": "What can you build with Amazon Bedrock AgentCore", "publisher": "AWS Documentation", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents AgentCore services working with any framework and model, serverless execution with isolated sessions, persistent memory, built-in observability, gateway wrapping of Lambda/OpenAPI into MCP tools, Harness orchestration, and a governed internal developer path with centralized authentication." }, { "id": "s2", "url": "https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/identity-overview.html", "title": "Overview of Amazon Bedrock AgentCore Identity", "publisher": "AWS Documentation", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents AgentCore Identity implementing authentication and authorization controls that verify each request independently and require explicit verification for all access attempts." } ], "claims": [ { "id": "c1", "statement": "Agents built with frameworks including CrewAI, LangGraph, LlamaIndex and Strands Agents can be deployed with serverless execution, isolated sessions, persistent memory and built-in observability, without infrastructure management.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The page does not state GA versus preview status for individual AgentCore services.", "capability": "framework-agnostic-agent-runtime" }, { "id": "c2", "statement": "Existing APIs, databases and services can be turned into tools for MCP-compatible agents, including a gateway that wraps Lambda functions or OpenAPI specifications without rewriting code.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Tool-level authorization granularity is not described on the overview page.", "capability": "gateway-tool-exposure" }, { "id": "c3", "statement": "AgentCore Identity implements authentication and authorization controls that verify each request independently and require explicit verification for all access attempts regardless of source.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The overview page does not document RBAC models, audit logging or human approval flows for agent actions.", "capability": "identity-and-authorization" }, { "id": "c4", "statement": "Provides internal developers a governed path to build and deploy agents using approved tools, shared memory stores and centralized observability, authentication and compliance.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "'Approved tools' and 'compliance' are stated at a high level without describing the approval or attestation mechanism.", "capability": "governed-internal-developer-path" }, { "id": "c5", "statement": "A built-in human-in-the-loop approval gate for agent actions is not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1", "s2" ], "limitations": "Absence of evidence; other AgentCore documentation pages were not fetched.", "capability": "human-approval-gate" } ], "limitations": [ "No availability status per service is stated in the overview, so GA versus preview boundaries are unclear.", "Approval workflows, agent-action audit trails and permission-scoping models were not evidenced in the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept as one entry covering AgentCore Runtime, Gateway, Identity, Memory and Observability, which AWS documents as composable parts of a single service family." }, { "name": "Amazon Bedrock Guardrails", "id": "amazon-bedrock-guardrails", "companyName": "Amazon Web Services", "companyUrl": "https://aws.amazon.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "runtime-controls", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Cloud security architect", "ML platform owner", "Compliance lead" ], "problemsSolved": [ "blocking prompt attacks and harmful content", "masking sensitive data in prompts and responses", "consistent safety policy across models" ], "sources": [ { "id": "s1", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html", "title": "Detect and filter harmful content by using Amazon Bedrock Guardrails", "publisher": "Amazon Web Services", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes configurable safeguards with content filters including Prompt Attack, sensitive information filters, use during inference by guardrail ID/version and via the ApplyGuardrail API." }, { "id": "s2", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails-how.html", "title": "How Amazon Bedrock Guardrails works", "publisher": "Amazon Web Services", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents evaluation of user inputs and model responses in parallel across policies, blocked messaging and masking on intervention, and support for Bedrock Agents and Knowledge Bases." } ], "claims": [ { "id": "c1", "statement": "Guardrails evaluate both user inputs and model responses against configured policies including content filters with a Prompt Attack category, denied topics, sensitive information filters and word filters, returning blocked messaging or masked content on intervention.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Documentation does not describe detection of tool misuse, agent action authorization or multi-step agent behaviour analysis.", "capability": "input-output-policy-evaluation" }, { "id": "c2", "statement": "Guardrails can be applied during foundation model inference by specifying a guardrail ID and version, or invoked independently through the ApplyGuardrail API without calling a model, and are supported with Bedrock Agents and Knowledge Bases.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Cross-account and cross-region behaviour is documented separately and was not verified here.", "capability": "flexible-invocation" } ], "limitations": [ "Content and data filtering only; no tool authorization", "Agent-specific threat detection not documented" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global", "US", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included as an LLM guardrail control with documented agent-adjacent support (Bedrock Agents); not a standalone agent threat detection platform." }, { "name": "Amazon SageMaker Model Registry", "id": "sagemaker-model-registry", "companyName": "Amazon Web Services", "companyUrl": "https://aws.amazon.com/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "ML platform engineer", "MLOps lead", "Model risk reviewer" ], "problemsSolved": [ "uncatalogued production models", "no record of model approval status", "missing model documentation for reviews" ], "sources": [ { "id": "s1", "url": "https://docs.aws.amazon.com/sagemaker/latest/dg/model-registry.html", "title": "Model Registration Deployment with Model Registry", "publisher": "Amazon Web Services", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents cataloging models, version management, metadata association, viewing model cards in registered models, lineage, staging construct, approval status management, CI/CD deployment, model groups and collections." }, { "id": "s2", "url": "https://docs.aws.amazon.com/sagemaker/latest/dg/model-cards.html", "title": "Amazon SageMaker Model Cards", "publisher": "Amazon Web Services", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents model card fields (intended use, risk rating, training details and metrics, evaluation results, considerations, recommendations), Model Registry integration, PDF export and immutable versioning on edit." } ], "claims": [ { "id": "c1", "statement": "AWS documentation states the Model Registry catalogs models for production, manages model versions, associates metadata such as training metrics, views model lineage for traceability and reproducibility, defines a staging construct for the model lifecycle, manages a model's approval status and automates deployment with CI/CD.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs describe an approval status field and staging, not a multi-party approval workflow with segregation of duties; scope is models registered in SageMaker.", "capability": "model-registry-versions-approval-lineage" }, { "id": "c2", "statement": "AWS documentation states model cards document intended use, risk rating, training details and metrics, evaluation results, considerations and recommendations in one place, are integrated with the Model Registry for auditing information, can be exported to PDF, and that any edit other than an approval-status update creates an additional card version forming an immutable record of changes.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The risk rating is a user-entered field; documentation does not validate content quality or satisfy any named regulation.", "capability": "model-cards-documentation-versioning" } ], "limitations": [ "Only AWS-hosted/registered models are in scope; governance of external AI is not addressed.", "No control-to-regulation mapping or evidence pack features documented." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Model Registry and Model Cards combined into one entry because AWS documents them as an integrated ML governance pair." }, { "name": "AppOmni Agent Inventory", "id": "appomni-agent-inventory", "companyName": "AppOmni", "companyUrl": "https://appomni.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "agent-security", "identity-access" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "SaaS security engineer", "GRC lead" ], "problemsSolved": [ "inventory of agents inside SaaS platforms", "over-permissioned agent detection", "unreviewed AI feature enablement" ], "sources": [ { "id": "s1", "url": "https://appomni.com/ai-security/agent-inventory/", "title": "Agent Inventory for SaaS AI Agents", "publisher": "AppOmni", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes surfacing agents across Agentforce, Now Assist and Copilot, capability mapping from declared tools, and identification of over-permissioned agents." }, { "id": "s2", "url": "https://appomni.com/press-releases/agentic-ai-security-for-servicenow-advances-aispm/", "title": "AppOmni Advances AISPM with Agentic AI Security for ServiceNow", "publisher": "AppOmni", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-11-19", "supports": "Announces AgentGuard for ServiceNow Now Assist agents covering discovery, posture, detection and response." } ], "claims": [ { "id": "c1", "statement": "Surfaces AI agents running in SaaS environments including Salesforce Agentforce, ServiceNow Now Assist and Microsoft 365 Copilot, including agents enabled without security review or internal approval.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; does not state the API mechanism per platform or how completely each platform exposes agent metadata.", "capability": "saas-agent-inventory" }, { "id": "c2", "statement": "Surfaces the tools and descriptions declared for each agent along with identities, permissions and privilege, and flags agents with excessive privileges.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Vendor notes visibility 'starts with what's known and documented', implying reliance on declared metadata rather than observed behaviour.", "capability": "capability-and-privilege-mapping" }, { "id": "c3", "statement": "AppOmni announced AgentGuard on 19 November 2025, extending the platform from agent discovery and posture to real-time prompt-injection prevention and DLP enforcement for ServiceNow Now Assist agents.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Runtime protection announced for ServiceNow only; other platforms are discovery/posture scope in this evidence.", "capability": "servicenow-agent-protection" } ], "limitations": [ "Only agents inside SaaS platforms AppOmni connects to are discoverable", "Behavioural capability discovery beyond declared metadata is not evidenced" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "ServiceNow Now Assist", "type": "native", "sourceIds": [ "s2" ], "limitations": "Press release describes agent-level protection for ServiceNow; depth for other SaaS platforms not established." } ], "conflicts": [], "editorialNotes": "Chosen over broader AppOmni SSPM marketing because Agent Inventory is the agent-specific product surface." }, { "name": "Arcade", "id": "arcade-actions-runtime", "companyName": "Arcade AI", "companyUrl": "https://www.arcade.dev/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "identity-access", "agent-building" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "Application security lead", "Platform engineer", "IAM manager" ], "problemsSolved": [ "delegated OAuth for agent tool calls", "per-action authorization at runtime", "blocking sensitive agent actions" ], "sources": [ { "id": "s1", "url": "https://docs.arcade.dev/home", "title": "Arcade Docs", "publisher": "Arcade", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Arcade as an enterprise actions runtime that handles OAuth and user tokens, enforces security policies per action, executes agent-optimised tools across MCP servers and governs agents centrally." }, { "id": "s2", "url": "https://www.arcade.dev/", "title": "Arcade: The Actions Runtime for Enterprise AI Agents", "publisher": "Arcade", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States agents act as the authenticated user within their own scope, supports pre- and post-tool-call hooks to block sensitive actions and redact PII, with cloud, on-prem, air-gapped or hybrid deployment." } ], "claims": [ { "id": "c1", "statement": "Arcade handles OAuth and user token management and enforces security policies on every action, adding per-action authorization at runtime on top of existing IdP, DLP and compliance policies.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation home page does not specify policy syntax or how conflicts with IdP policy are resolved.", "capability": "runtime-action-authorization" }, { "id": "c2", "statement": "The product page states pre- and post-tool-call hooks can inspect requests and responses to block sensitive actions, redact PII and prevent data leaving systems.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Hook behaviour and failure modes are asserted on the product page rather than documented in reference material fetched here.", "capability": "tool-call-hooks" }, { "id": "c3", "statement": "Arcade provides a large catalogue of agent-optimised tools across MCP servers and supports connection from MCP clients such as Cursor, VS Code and Claude Desktop, with a central governance control plane.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Tool counts are vendor-reported; individual tool integrations were not verified.", "capability": "tool-catalog-and-mcp" } ], "limitations": [ "Some enforcement detail only on marketing pages", "Open-source component is the tool framework, not the runtime" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Categorised primarily as runtime authorization because the evidenced differentiator is per-action authorization and credential brokering, not agent authoring." }, { "name": "Arize AX", "id": "arize-ax", "companyName": "Arize AI", "companyUrl": "https://arize.com", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "ML platform lead", "AI engineer" ], "problemsSolved": [ "agent trace and trajectory debugging", "online and offline evaluation", "failure clustering and issue triage", "dataset and experiment management" ], "sources": [ { "id": "s1", "url": "https://arize.com/docs/ax", "title": "Arize AX Docs", "publisher": "Arize AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes capturing traces of real behaviour, recording each run with inputs, outputs, tools and costs, 30+ provider and framework integrations, annotations, evaluators including LLM/agent-as-a-judge and code evaluators, dataset experiments and ranked issues with trace evidence." }, { "id": "s2", "url": "https://arize.com/pricing/", "title": "Arize AX Pricing", "publisher": "Arize AI", "type": "official_pricing", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists trace-span and ingestion allowances per plan, session support, multi-modal tracing, agent trajectory visualisations, online and offline evals, trace and session evals, agent-as-judge, human annotation, and describes Phoenix as the open-source local-first platform." } ], "claims": [ { "id": "c1", "statement": "Instrumentation records each agent run as traces and spans capturing inputs, outputs, tools and costs, with pre-built integrations for 30+ providers and frameworks.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs overview does not describe multi-agent handoff semantics or retention of captured spans.", "capability": "trace-capture" }, { "id": "c2", "statement": "Documented evaluation features include continuous evaluator labelling of traces, LLM-as-a-judge, agent-as-a-judge and code evaluators, dataset-based experiments, human annotation of spans, and evaluator alignment against human judgment.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Evaluator accuracy is not established by the vendor pages.", "capability": "evaluation-and-annotation" }, { "id": "c3", "statement": "Pricing documents session support, multi-modal tracing for image, voice and PDF, agent trajectory path and graph visualisations, trace evals for agent trajectories, session evals for multi-turn conversations, and metered trace-span and ingestion limits per plan.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Plan feature lists do not state audit logging, retention configuration or regional hosting options.", "capability": "session-and-trajectory-features" }, { "id": "c4", "statement": "Arize maintains Phoenix, described on the pricing page as an open-source, local-first platform for tracing, evaluation, experimentation and prompt iteration.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Phoenix licensing and feature parity with AX were not verified from a repository page in this session.", "capability": "open-source-companion" } ], "limitations": [ "A dedicated security and compliance docs page returned an error, so audit logging, retention and access control remain unverified.", "Governance value is indirect: rich telemetry and evaluation, but no documented immutable audit record or policy gate." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": "tiered plans (AX Free, AX Pro, AX Enterprise) metered on trace spans and ingestion volume", "url": "https://arize.com/pricing/" }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "One entry for Arize AX; Phoenix recorded as the vendor's open-source companion rather than a separate entry because only the pricing-page description was verified. https://arize.com/docs/ax/security-and-compliance returned HTTP client error." }, { "name": "Asenion", "id": "asenion", "companyName": "Asenion", "companyUrl": "https://www.asenion.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief risk officer", "AI governance lead", "Model validator" ], "problemsSolved": [ "manual AI control testing", "EU AI Act readiness for predictive and generative AI", "evidence of ongoing AI oversight" ], "sources": [ { "id": "s1", "url": "https://www.asenion.ai/", "title": "Asenion | AI Governance, Risk and Compliance Management", "publisher": "Asenion", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes AI TRiSM scope, lifecycle oversight of AI systems and agents, automated controls that continuously assess and test, verifiable tamper-resistant assurance, and coverage beyond EU AI Act, ISO/IEC 42001 and NIST AI RMF." }, { "id": "s2", "url": "https://www.asenion.ai/blog/fairly-ai-acquires-anch-ai-to-create-asenion", "title": "Fairly AI acquires Anch.AI to launch Asenion", "publisher": "Asenion", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-06-18", "supports": "States Fairly AI (Kitchener-Waterloo, Canada) acquired Swedish anch.AI, new company named Asenion, platform made available, covers predictive/generative/agentic AI, includes testing agents for fairness, privacy and security." }, { "id": "s3", "url": "https://www.fairly.ai/", "title": "Fairly AI", "publisher": "Fairly AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Legacy Fairly AI site still live, marketing a fast track to the EU AI Act." } ], "claims": [ { "id": "c1", "statement": "Company site states Asenion provides automated controls that continuously assess, test and govern AI systems and agents across the lifecycle, with verifiable, tamper-resistant assurance and coverage of regulations and standards including the EU AI Act, ISO/IEC 42001 and NIST AI RMF.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page; test methods, what makes records tamper-resistant, and framework coverage detail are not evidenced, and no registry/approval features are described.", "capability": "automated-controls-and-assurance" }, { "id": "c2", "statement": "A company post dated June 18, 2025 states Fairly AI acquired Swedish start-up anch.AI to launch Asenion as the new company name, covering predictive, generative and agentic AI, with testing agents for fairness, privacy and security, and that the platform has been made available.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Vendor announcement; deal terms and current product continuity between Fairly AI and anch.AI codebases are not established.", "capability": "acquisition-and-rebrand" }, { "id": "c3", "statement": "The legacy fairly.ai domain still resolves and markets a fast track to the EU AI Act, indicating the older brand remains partially online after the rebrand.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Only confirms the page exists; it does not establish which brand is actively sold or supported.", "capability": "prior-brand-pages-live" } ], "limitations": [ "No technical documentation located for either brand.", "Unclear whether anch.AI's assessment content has been fully merged into Asenion." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "CA", "SE" ], "pricing": { "model": null, "url": null }, "formerNames": [ "Fairly AI", "anch.AI", "AI Sustainability Center" ], "parentCompany": null, "integrations": [], "conflicts": [ { "issue": "Brand coexistence", "detail": "asenion.ai presents Asenion as the company post-acquisition while fairly.ai remains live with Fairly AI branding.", "sourceIds": [ "s1", "s3" ] } ], "editorialNotes": "Dedup decision: Fairly AI and anch.AI merged into a single Asenion entry; formerNames records both." }, { "name": "Astrix Agent Control Plane", "id": "astrix-agent-control-plane", "companyName": "Astrix Security", "companyUrl": "https://astrix.security", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "agent-security" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Identity security lead", "SOC manager" ], "problemsSolved": [ "shadow agent and NHI discovery", "owner attribution for agents and secrets", "policy enforcement and offboarding for agents" ], "sources": [ { "id": "s1", "url": "https://astrix.security/product/ai-agent-discovery/", "title": "AI Agent Discovery & Visibility Platform | Astrix", "publisher": "Astrix Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-03-30", "supports": "Details discovery of agents, MCP servers, NHIs and secrets, and the identity graph mapping owners, permissions and reachable resources." }, { "id": "s2", "url": "https://astrix.security/product/secure-ai-agents/", "title": "Secure AI Agents & NHIs at Scale | Astrix Security", "publisher": "Astrix Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the Agent Control Plane, agent onboarding/offboarding, owner assignment, policy rules and behavioural anomaly detection." }, { "id": "s3", "url": "https://www.prnewswire.com/news-releases/astrix-security-delivers-the-most-comprehensive-ai-agent-discovery-and-enhances-security-with-agent-policy-enforcement-302719653.html", "title": "Astrix Security Delivers the Most Comprehensive AI Agent Discovery and Enhances Security with Agent Policy Enforcement", "publisher": "PR Newswire", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-03-23", "supports": "Announces Agent Policies, agentic just-in-time access and four discovery methods including AI platform integrations." } ], "claims": [ { "id": "c1", "statement": "Astrix automatically discovers custom, third-party, shadow and managed AI agents, MCP servers, NHIs (service accounts, OAuth apps, API keys, SSH keys, IAM roles) and secrets inside and outside vaults, and maps them in an identity graph with owners.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Product page does not list the specific connectors or their coverage limits.", "capability": "agent-and-nhi-discovery" }, { "id": "c2", "statement": "The platform can onboard shadow agents, offboard unused agents and agents of departing employees, revoke access after policy violations and automate secret rotation.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Automation scope per integrated system is not documented on the fetched page.", "capability": "agent-lifecycle-actions" }, { "id": "c3", "statement": "Astrix released Agent Policies and agentic just-in-time access alongside four AI agent discovery methods in March 2026.", "support": "independently_supported", "sourceIds": [ "s3" ], "limitations": "Vendor-issued press release distributed by PR Newswire; no third-party validation of coverage claims.", "capability": "agent-policy-enforcement" } ], "limitations": [ "No fetched source states pricing, deployment options beyond SaaS, or certification status." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Agent Control Plane (ACP) announced Sep 2025 per vendor page; treated as the agent-facing product within the Astrix platform." }, { "name": "Auth0 Auth for GenAI (Token Vault & delegated authorization)", "id": "auth0-auth-for-genai", "companyName": "Auth0 (Okta)", "companyUrl": "https://auth0.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "unknown", "buyerRoles": [ "Application security engineer", "Platform engineer", "Product engineering lead" ], "problemsSolved": [ "agent acting on behalf of a user", "keeping third-party API credentials out of agent code", "scoping agent API access to user consent" ], "sources": [ { "id": "s1", "url": "https://auth0.com/ai/docs/intro/delegated-authorization", "title": "Delegated Authorization for AI Agents", "publisher": "Auth0", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes OAuth-based delegated authorization for agents and points to Token Vault for third-party API access." }, { "id": "s2", "url": "https://auth0.com/ai/docs/intro/token-vault", "title": "Using Token Vault", "publisher": "Auth0", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Defines Token Vault as a service for storing and managing tokens for external services that agents fetch to call APIs on a user's behalf." } ], "claims": [ { "id": "c1", "statement": "Auth0 issues access tokens that scope an AI agent's API calls to the permissions a user explicitly consented to, for first-party and third-party APIs.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs do not describe autonomous (no-user) agent identities or agent inventory.", "capability": "on-behalf-of-user-delegation" }, { "id": "c2", "statement": "Token Vault obtains, stores and refreshes third-party API tokens so the agent never handles those credentials directly.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "No documented list of guaranteed connectors or token retention controls on the fetched pages.", "capability": "external-token-vaulting" } ], "limitations": [ "No fetched source states availability stage or pricing." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": "Okta", "integrations": [], "conflicts": [], "editorialNotes": "Product family is marketed as 'Auth for GenAI'; entry is scoped to the evidenced delegated-authorization and Token Vault capabilities." }, { "name": "Azure AI Content Safety Prompt Shields", "id": "azure-prompt-shields", "companyName": "Microsoft", "companyUrl": "https://azure.microsoft.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI application developer", "Cloud security architect", "Responsible AI lead" ], "problemsSolved": [ "direct prompt attack detection", "indirect injection detection in documents", "reusable safety API for AI apps" ], "sources": [ { "id": "s1", "url": "https://learn.microsoft.com/en-us/azure/ai-services/content-safety/concepts/jailbreak-detection", "title": "Prompt Shields in Azure AI Content Safety", "publisher": "Microsoft", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Prompt Shields as a unified API detecting and blocking adversarial user input attacks, including user prompt injection and document attacks with embedded hidden instructions; notes the former name Jailbreak risk detection." }, { "id": "s2", "url": "https://learn.microsoft.com/en-us/azure/ai-services/content-safety/quickstart-jailbreak", "title": "Quickstart: Detect prompt attacks with Prompt Shields", "publisher": "Microsoft", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the shieldPrompt API version 2024-09-01, detection of user prompt attacks and document attacks, attackDetected response flags and adjustable thresholds." } ], "claims": [ { "id": "c1", "statement": "Prompt Shields is a unified API that detects and blocks adversarial user input attacks on LLMs, covering user prompt injection attempts to circumvent system rules and document-based indirect attacks with hidden instructions in external material.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Detection is scoped to prompts and documents; tool calls, agent plans and outbound data flows are not covered.", "capability": "direct-and-indirect-attack-detection" }, { "id": "c2", "statement": "The API endpoint contentsafety/text:shieldPrompt returns attackDetected booleans for the user prompt and each document, with configurable thresholds for filtering.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Enforcement depends on the calling application; the API itself does not block traffic.", "capability": "api-response-and-thresholds" } ], "limitations": [ "Detection API only, no enforcement or agent action control", "No tool-call or exfiltration analysis" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global", "US", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [ "Jailbreak risk detection" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Distinct from Azure AI Foundry agent governance features; scoped here to the Prompt Shields detection API." }, { "name": "BABL AI audits and AI & Algorithm Auditor Certification", "id": "babl-ai", "companyName": "BABL AI", "companyUrl": "https://babl.ai", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "evaluation" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Head of Compliance", "AI governance lead", "Internal auditor" ], "problemsSolved": [ "independent algorithm audit", "AI risk and impact assessment", "auditor upskilling" ], "sources": [ { "id": "s1", "url": "https://babl.ai/", "title": "BABL AI - We Audit Algorithms", "publisher": "BABL AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes AI audits, responsible AI consulting, independent third-party audits and certified independent auditors." }, { "id": "s2", "url": "https://courses.babl.ai/p/ai-and-algorithm-auditor-certification", "title": "AI and Algorithm Auditor Certification", "publisher": "BABL AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Details the five-course structure, capstone and exit exam and the target audiences." } ], "claims": [ { "id": "c1", "statement": "BABL AI states it performs independent third-party audits and AI audit and compliance work, employs certified independent auditors, and that its audits follow globally recognised assurance engagement standards similar to financial auditing.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No specific assurance standard is named, and no accreditation or oversight body for BABL's auditors is evidenced.", "capability": "independent-audits" }, { "id": "c2", "statement": "BABL AI's certification program consists of five core courses plus a capstone project and exit exam, aimed at auditors, AI governance specialists and risk professionals who perform AI risk and impact assessments.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "It is a vendor-issued certificate; no external accreditation, ANSI/ISO 17024-style recognition, or regulator acceptance is evidenced.", "capability": "auditor-certification-program" } ], "limitations": [ "Assurance standard referenced only generically", "Credential is vendor-issued with no evidenced external accreditation" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "US", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Audits and the training credential kept in one entry with separate claims; buyers differ (organisations vs individual practitioners)." }, { "name": "BigID AI Security & Governance", "id": "bigid-ai-security-governance", "companyName": "BigID", "companyUrl": "https://bigid.com", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "agent-discovery", "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Chief data officer", "Privacy counsel" ], "problemsSolved": [ "AI asset inventory", "sensitive data classification for AI", "AI data lineage" ], "sources": [ { "id": "s1", "url": "https://bigid.com/ai-security-governance/", "title": "AI Security & Governance Platform - BigID", "publisher": "BigID", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-08-13", "supports": "Describes AI system discovery, AI asset inventory covering agents and vector stores, classification of data used in training/retrieval/inference, and AI data lineage." }, { "id": "s2", "url": "https://bigid.com/", "title": "BigID: Enterprise Data Security Platform for DSPM & AI", "publisher": "BigID", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-05-08", "supports": "Positions the Data Security Platform combining DSPM, DLP, access, privacy and labeling; states scoped RBAC and ability to track and restrict genAI data access by user, model or agent, with audit trails for rights workflows." } ], "claims": [ { "id": "c1", "statement": "Discovers and inventories AI models, agents, copilots, applications, datasets, prompts, pipelines, vector databases and shadow AI across the enterprise.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; no documentation of scanners, supported platforms or detection method per asset type.", "capability": "ai-asset-discovery" }, { "id": "c2", "statement": "Maps how sensitive data flows through AI training, tuning, inference, retrieval, prompts and downstream workflows.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Lineage granularity and whether it is derived from scanning or from platform metadata is not stated.", "capability": "ai-data-lineage" }, { "id": "c3", "statement": "Provides scoped role-based access control by person, team or domain and the ability to track and restrict access to genAI data by user, model or agent.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Homepage marketing text; enforcement point for restricting agent access to data is not documented.", "capability": "scoped-rbac-and-genai-access-control" } ], "limitations": [ "Documentation portal was not reachable, so all claims rest on vendor product pages.", "It is unclear which capabilities belong to the BigID Next generation versus the older platform." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "IL", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included because the vendor pages describe AI-specific asset discovery and AI data lineage, not merely AI-flavoured DLP marketing. 'BigID Next' appears in third-party listings but was not confirmed as a distinct SKU." }, { "name": "Boomi Agent Control Tower", "id": "boomi-agent-control-tower", "companyName": "Boomi", "companyUrl": "https://boomi.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "agent-building", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Integration platform owner", "Enterprise architect", "IT operations manager" ], "problemsSolved": [ "cross-provider agent registry", "agent sprawl control", "agent activity monitoring" ], "sources": [ { "id": "s1", "url": "https://help.boomi.com/docs/Atomsphere/Platform/Agent_Control_Tower", "title": "Boomi Agent Control Tower", "publisher": "Boomi", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents registering and managing agents across providers, metadata on goals/tasks/tools, trust levels and tags, monitoring, and provider-level enable/disable." }, { "id": "s2", "url": "https://boomi.com/blog/agent-control-tower-agentic-governance/", "title": "Conquer Agent Sprawl With Boomi Agent Control Tower", "publisher": "Boomi", "type": "other", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Vendor blog framing Agent Control Tower as governance for agent sprawl alongside data pipelines feeding agents." } ], "claims": [ { "id": "c1", "statement": "Lets administrators register, manage, govern and monitor agents across environments from one location, covering agents built in Boomi Agentstudio and third-party providers such as Amazon Bedrock.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation names Boomi Agent Garden and Amazon Bedrock as provider examples; the full supported provider list is on a separate page not reviewed.", "capability": "cross-provider-agent-registry" }, { "id": "c2", "statement": "Provides metadata on agent goals, tasks and connected tools, allows trust levels and tags for organisation, supports monitoring of usage and errors, and for some providers allows enabling or disabling an agent directly.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Enable/disable is provider-dependent; documentation does not state which providers support it beyond Bedrock.", "capability": "agent-metadata-and-controls" }, { "id": "c3", "statement": "Boomi positions Agent Control Tower as the governance response to agent sprawl created by low-code and vibe-coded agent creation.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Vendor blog framing; no independent verification of governance outcomes.", "capability": "agent-sprawl-positioning" } ], "limitations": [ "Registry depends on connected provider accounts; no detection of unmanaged agents", "Full provider list not verified this session" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Amazon Bedrock", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Documented as a connectable provider with enable/disable support; scope of imported metadata not fully specified." } ], "conflicts": [], "editorialNotes": "Primary category is discovery/inventory because the documented function is a registry; Agentstudio's build features are secondary." }, { "name": "Braintrust", "id": "braintrust", "companyName": "Braintrust", "companyUrl": "https://www.braintrust.dev", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI engineering lead", "Head of ML platform" ], "problemsSolved": [ "prompt and agent regression testing", "dataset curation from production logs", "quality scoring of AI output" ], "sources": [ { "id": "s1", "url": "https://www.braintrust.dev/docs", "title": "Get started with Braintrust", "publisher": "Braintrust", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes logging traces, analyzing logs, adding human feedback and building datasets, running evals and experiments, monitoring production, and managing organizations, projects and access control." }, { "id": "s2", "url": "https://www.braintrust.dev/docs/guides/self-hosting", "title": "Self-hosting Braintrust", "publisher": "Braintrust", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the self-hosted data plane on AWS/GCP/Azure with customer IAM policies, KMS keys and audit trails; Braintrust retains the control plane with UI, authentication and user management; infra dashboard limited to owners and Manage settings permission." } ], "claims": [ { "id": "c1", "statement": "Braintrust documents capturing traces from an AI application, analyzing logs, adding human feedback to build datasets, and running evals and experiments against them.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The quickstart does not enumerate span types (LLM call, tool call, multi-agent handoff) or describe retention of logged data.", "capability": "trace-logging-and-evals" }, { "id": "c2", "statement": "A self-hosted deployment option lets the customer run the data plane in their own AWS, GCP or Azure account using their own IAM policies, KMS keys and audit trails, while Braintrust hosts the control plane, UI and authentication.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The audit trails referenced are the customer's own cloud-provider trails, not an in-product immutable audit log of evaluation activity.", "capability": "self-hosted-data-plane" }, { "id": "c3", "statement": "Access to organization, project and infrastructure settings is governed by permission groups, with the infra dashboard limited to organization owners and members holding the Manage settings permission.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "No documented evidence for SSO enforcement, SCIM provisioning or per-record retention policies.", "capability": "access-control" } ], "limitations": [ "Primarily a developer quality-engineering tool; no documented immutable audit record, retention policy or policy-gating control for governance reporting.", "Vendor now markets the product as an 'active observability platform', so the evaluation/observability boundary is blurred." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Classified evaluation-primary because docs lead with evals, experiments, scorers and datasets; observability added as secondary since the same docs cover trace logging and production monitoring." }, { "name": "Britive ARC (Agentic Runtime Control)", "id": "britive-arc", "companyName": "Britive", "companyUrl": "https://www.britive.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Cloud security architect", "IAM manager", "Platform engineer" ], "problemsSolved": [ "standing privileges for agents and NHIs", "agent onboarding and retirement", "controlling agent access to MCP tools" ], "sources": [ { "id": "s1", "url": "https://www.britive.com/platform/agentic-ai-identity-security", "title": "Agentic Identity Security: Runtime Access for AI Identities - Britive", "publisher": "Britive", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Britive ARC agent registration, lifecycle governance, standards-based agent authentication and per-task credential creation and destruction." }, { "id": "s2", "url": "https://www.britive.com/resource/blog/agentic-ai-identity-security-capability-launch", "title": "Britive launches Agentic AI Identity Security capabilities", "publisher": "Britive", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Launch post (stated September 2025) covering agent identity lifecycle management, runtime authorization with zero standing privileges and agent-to-tool MCP access." } ], "claims": [ { "id": "c1", "statement": "Every agent is registered as its own identity with a named human owner, job function, business use and allowed access profiles, governed from onboarding through execution to removal.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Vendor platform and blog pages; no reference documentation fetched.", "capability": "agent-registration-and-lifecycle" }, { "id": "c2", "statement": "Where a target requires a credential, Britive creates it at request time and destroys it when the task ends, so the agent never holds a privileged credential between tasks.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Supported targets for permission elevation versus credential creation are not enumerated.", "capability": "ephemeral-privilege-issuance" }, { "id": "c3", "statement": "Agents authenticate using SPIFFE SVID, OIDC federation or API tokens, resolved on every request.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documentation of attestation depth or replay protections.", "capability": "standards-based-agent-auth" } ], "limitations": [ "Publication dates on Britive pages were inconsistent with the stated September 2025 launch; treated as unknown." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "issue": "Page metadata dates (2024-07-23) conflict with the in-text September 2025 launch statement.", "sourceIds": [ "s2" ] } ], "editorialNotes": "ARC is the branded agentic runtime control within the Britive cloud PAM platform." }, { "name": "BSI ISO/IEC 42001 certification", "id": "bsi-iso-42001-certification", "companyName": "BSI", "companyUrl": "https://www.bsigroup.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "model-governance" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Head of Compliance", "Quality manager", "Chief AI Officer" ], "problemsSolved": [ "ISO/IEC 42001 certification", "AI management system gap assessment", "auditor and implementer training" ], "sources": [ { "id": "s1", "url": "https://www.bsigroup.com/en-GB/insights-and-media/media-centre/press-releases/2025/november/bsi-becomes-the-first-certification-body-accredited-by-ukas-and-rva-to-deliver-certification-for-isoiec-42001/", "title": "BSI first to gain UKAS and RvA accreditation to certify ISO/IEC 42001", "publisher": "BSI", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States UKAS and RvA accreditation for ISO/IEC 42001 certification and that EU AI Act notified body designation is still under evaluation." }, { "id": "s2", "url": "https://www.bsigroup.com/en-US/products-and-services/standards/iso-42001-ai-management-system/", "title": "ISO 42001 - AI Management System", "publisher": "BSI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States triple UKAS/RvA/ANAB accreditation and lists certification, pre-certification assessment and training." } ], "claims": [ { "id": "c1", "statement": "BSI states it was the first certification body accredited by UKAS to certify ISO/IEC 42001 and that it also holds RvA (Netherlands) and ANAB accreditation for certifying AI management systems.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Accreditation is stated by BSI itself, not read from UKAS/RvA/ANAB registers in this session; the accreditation covers ISO/IEC 42001 management-system certification only.", "capability": "accredited-scheme" }, { "id": "c2", "statement": "BSI's release states it is undergoing evaluation for other accredited services such as accreditation and designation as an EU AI Act notified body, i.e. it does not claim that designation today.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Status is 'undergoing evaluation'; no designation has been evidenced and no scope under the EU AI Act can be assumed.", "capability": "eu-ai-act-notified-body-status-pending" }, { "id": "c3", "statement": "BSI's product page lists ISO/IEC 42001 certification, a Pre-Certification Assessment and ISO/IEC 42001 training courses.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Page does not disclose audit duration, fees, or auditor qualification criteria.", "capability": "supporting-services" } ], "limitations": [ "Accreditation confirmed only from BSI's own pages, not accreditation-body registers", "No EU AI Act notified-body designation evidenced" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "UK", "NL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Accreditation claims deliberately quoted narrowly: scope is ISO/IEC 42001 AI management systems certification." }, { "name": "Camunda Agentic Orchestration", "id": "camunda-agentic-orchestration", "companyName": "Camunda", "companyUrl": "https://camunda.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "work-delegation" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "Process automation lead", "Enterprise architect", "Engineering manager" ], "problemsSolved": [ "embedding agents in end-to-end business processes", "mixing deterministic rules with agent decisions", "coordinating humans and agents" ], "sources": [ { "id": "s1", "url": "https://docs.camunda.io/docs/components/agentic-orchestration/agentic-orchestration-overview/", "title": "Agentic orchestration - Camunda 8 Docs", "publisher": "Camunda", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents Camunda agentic orchestration in version 8.9, with agents executing non-deterministic parts of BPMN processes alongside human tasks and deterministic rules, agent memory, evaluation and Camunda SaaS with a Camunda-provided LLM." }, { "id": "s2", "url": "https://docs.camunda.io/docs/components/agentic-orchestration/ai-agents/", "title": "AI agents | Camunda 8 Docs", "publisher": "Camunda", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents LLM provider support, exposure of BPMN activities inside ad-hoc sub-processes as LLM-selectable tools, and routing of human tasks and events." }, { "id": "s3", "url": "https://docs.camunda.io/docs/components/concepts/access-control/authorizations/", "title": "Orchestration Cluster authorization - Camunda 8 Docs", "publisher": "Camunda", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the fine-grained authorization system with owners (User, Group, Role, Client, Mapping rule), resource permissions including user-task read/update on process definitions, Tenant as a resource, and READ_USAGE_METRIC on System." } ], "claims": [ { "id": "c1", "statement": "AI agents execute the non-deterministic parts of a process while human tasks, deterministic rule sets and AI-driven decisions collaborate in one BPMN-based end-to-end process.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The overview does not document audit logs, RBAC, versioning, tenant isolation or observability for agent behaviour.", "capability": "bpmn-agent-orchestration" }, { "id": "c2", "statement": "BPMN activities inside an ad-hoc sub-process are exposed as tools the LLM can select, and agents connect to providers including Anthropic, Amazon Bedrock, Google Gemini and OpenAI; human tasks and events are routed within the process.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No approval-specific control is documented; human involvement is modelled as ordinary human tasks.", "capability": "bpmn-activities-as-tools" }, { "id": "c3", "statement": "The Camunda 8 Orchestration Cluster provides a fine-grained authorization system over web components and APIs, with owner types User, Group, Role, Client and Mapping rule, permissions such as READ_USER_TASK and UPDATE_USER_TASK on process definitions, and Tenant as an authorization resource.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Tenant appears as an authorization resource, but the page does not explicitly document tenant isolation guarantees; no audit-log capability is described.", "capability": "fine-grained-authorization" } ], "limitations": [ "No agent-specific audit log, approval gate or agent versioning documented; governance derives from generic process and cluster controls.", "Availability status of agentic orchestration features is not stated on the overview page." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "DE", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Entered as the agentic-orchestration capability of Camunda 8 rather than the whole BPM platform, per the AI-specific-evidence rule. productType hybrid: Camunda 8 has open-source components plus commercial SaaS and self-managed offerings (license not verified this session). Also surfaced in work-delegation research; agent-building kept as primary because the evidenced function is BPMN workflow orchestration, not governed duty assignment." }, { "name": "Check Point AI Agent Security (formerly Lakera Guard)", "id": "checkpoint-ai-agent-security", "companyName": "Check Point Software Technologies", "companyUrl": "https://www.checkpoint.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AI security lead", "Application security manager" ], "problemsSolved": [ "prompt injection detection in agent workflows", "restricting which tools an agent may call", "agent inventory and risk rating" ], "sources": [ { "id": "s1", "url": "https://docs.lakera.ai/guard", "title": "Deployment Options - Check Point AI Agent Security", "publisher": "Check Point Software Technologies", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Check Point AI Agent Security discovering agents, assessing configuration risk and protecting at runtime via AI Guardrails, with prompt injection detection across tool calls, responses and descriptions, Dangerous Deviation detection and Tool Allow/Deny lists." }, { "id": "s2", "url": "https://www.lakera.ai/lakera-guard", "title": "Agent Security", "publisher": "Lakera (Check Point)", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Product page states policy and runtime enforcement governs what agents can access, call and do, blocking prompt attacks, data leakage, unsafe tool use and unauthorised agent actions as agents operate." } ], "claims": [ { "id": "c1", "statement": "Documentation states detection of prompt injections, jailbreaks and manipulation in user prompts, reference materials, tool responses and tool descriptions across the agentic workflow.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Detection accuracy rates and supported languages are not stated on the fetched page.", "capability": "injection-detection-across-agent-surfaces" }, { "id": "c2", "statement": "A Tool Allow/Deny List controls which tools an agent may call, and a Dangerous Deviation detector flags actions outside the agent's trusted mandate.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Enforcement point and integration requirements for blocking are not fully specified on this page.", "capability": "tool-allow-deny-and-deviation" }, { "id": "c3", "statement": "The product connects to agent platforms and cloud infrastructure to build a continuously updated inventory of agents, their tools and connected MCP servers, with risk ratings mapped to OWASP and MITRE ATLAS.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Coverage is limited to the listed supported platforms.", "capability": "agent-discovery-inventory" }, { "id": "c4", "statement": "AI Guardrails is available as a standalone runtime layer for teams embedding the detection directly into their own AI applications.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Packaging, quotas and pricing are not documented.", "capability": "standalone-guardrails" } ], "limitations": [ "Product naming is in transition between Lakera and Check Point branding", "Discovery coverage limited to supported agent platforms" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "library" ], "geographies": [ "IL", "CH", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "Lakera Guard" ], "parentCompany": "Check Point Software Technologies", "integrations": [], "conflicts": [], "editorialNotes": "Lakera was acquired by Check Point; docs.lakera.ai now presents the product as Check Point AI Agent Security with AI Guardrails as the runtime layer. Recorded as one entry with Lakera Guard as a former name to avoid duplicate listings." }, { "name": "Checkmarx AI Inventory and AI-BOM", "id": "checkmarx-ai-inventory", "companyName": "Checkmarx", "companyUrl": "https://checkmarx.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "ai-risk", "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AppSec manager", "CISO", "Compliance lead" ], "problemsSolved": [ "AI component inventory in the SDLC", "AI-BOM generation for audits", "policy blocking of unapproved AI components" ], "sources": [ { "id": "s1", "url": "https://checkmarx.com/ai-bom/", "title": "AI-BOM for Enterprise AI Governance", "publisher": "Checkmarx", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States AI-BOM inventories models, agents, MCP servers and SDKs from source code and configuration on every commit, with PR/CI policy enforcement inside Checkmarx One." }, { "id": "s2", "url": "https://checkmarx.com/press-releases/checkmarx-delivers-complete-ai-asset-visibility-governance-software-supply-chain-closing-shadow-ai-gap/", "title": "Checkmarx Delivers Complete AI Asset Visibility and Governance", "publisher": "Checkmarx", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-06-23", "supports": "Announces general availability of Checkmarx AI Inventory in Checkmarx One with AI-BOM generation and policy controls for models, agents and MCP servers." } ], "claims": [ { "id": "c1", "statement": "Inventories models, agents, MCP servers, AI libraries and SDKs directly from source code and configuration files across repositories, refreshed on every commit.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; language and framework coverage of the detection engine is not enumerated.", "capability": "code-based-ai-discovery" }, { "id": "c2", "statement": "Checkmarx announced general availability of Checkmarx AI Inventory within Checkmarx One on 23 June 2026, describing deterministic detection and AI-BOM generation for discovered components.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Press release does not detail supported languages, agent frameworks, or false-positive handling.", "capability": "ga-availability" }, { "id": "c3", "statement": "Policies on approved models, agents, MCP servers and SDKs can be enforced in pull requests and CI/CD pipelines, flagging or blocking unapproved components.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documentation reviewed on policy syntax or enforcement points.", "capability": "pipeline-policy-enforcement" } ], "limitations": [ "Detects AI assets referenced in code; runtime-only or SaaS-side agents are outside this mechanism", "No public pricing reviewed" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "IL", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Marketing page uses the name AI-BOM; the press release names the GA capability AI Inventory. Treated as one product." }, { "name": "Cisco AI Defense AI BOM", "id": "cisco-ai-defense-aibom", "companyName": "Cisco", "companyUrl": "https://www.cisco.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "model-governance" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AppSec engineer", "Platform engineer", "AI security researcher" ], "problemsSolved": [ "AI-BOM generation from source", "agent and MCP server discovery in code", "container image AI component inventory" ], "sources": [ { "id": "s1", "url": "https://github.com/cisco-ai-defense/aibom", "title": "cisco-ai-defense/aibom", "publisher": "Cisco AI Defense (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "README states the tool scans codebases, container images and cloud environments to produce an AI-BOM of models, agents, tools, MCP servers/clients, datasets, prompts, guardrails and secrets across seven languages." }, { "id": "s2", "url": "https://cisco-ai-defense.github.io/docs/aibom", "title": "AI BOM | Cisco AI Defense", "publisher": "Cisco", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents libcst static analysis, DuckDB catalog matching, container image scanning, custom catalogs and inline annotations." } ], "claims": [ { "id": "c1", "statement": "Scans codebases, container images and cloud environments to produce an inventory of models, agents, tools, MCP servers and clients, datasets, prompts, guardrails and secrets, with support listed for Python, JavaScript/TypeScript, Java, Go, Rust, Ruby and C#.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Repository README describes capability; depth of non-Python language support and cloud scanning is not detailed there.", "capability": "multi-target-scanning" }, { "id": "c2", "statement": "Uses libcst-based static Python analysis to capture assignments, decorators, type annotations, context managers and class definitions, matching fully qualified symbols against a DuckDB catalog, with optional LLM enrichment for model names and call-path annotation.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documentation describes the Python analysis path specifically; detection is limited to components represented in the catalog or custom rules.", "capability": "static-analysis-method" }, { "id": "c3", "statement": "Supports custom component catalogs, base-class detection rules, exclude patterns and inline source annotations via a .aibom.yaml file and comment tags.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Requires engineering effort to tune; no managed UI or continuous monitoring is claimed.", "capability": "extensibility" } ], "limitations": [ "Point-in-time scanning tool, not a continuously updated enterprise inventory", "License and release cadence not assessed in detail" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "self_hosted" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included as the open-source representative for AI-BOM tooling because it has official Cisco repository and documentation evidence and explicitly covers agents and MCP servers." }, { "name": "Cisco MCP Scanner", "id": "cisco-mcp-scanner", "companyName": "Cisco", "companyUrl": "https://www.cisco.com", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "evaluation" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AppSec engineer", "CISO", "AI platform lead" ], "problemsSolved": [ "malicious MCP tool detection", "MCP supply-chain review", "behavioural code analysis of agent tools" ], "sources": [ { "id": "s1", "url": "https://github.com/cisco-ai-defense/mcp-scanner", "title": "cisco-ai-defense/mcp-scanner", "publisher": "Cisco / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "README describes the three scanning engines, source-code behavioural analysis, dependency auditing and binary hash lookups." }, { "id": "s2", "url": "https://api.github.com/repos/cisco-ai-defense/mcp-scanner", "title": "GitHub REST API repository record", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Reports Apache-2.0 license, archived=false and last push 2026-09-04." } ], "claims": [ { "id": "c1", "statement": "The repository states MCP Scanner combines the Cisco AI Defense inspect API, YARA rules and LLM-based analysis to detect malicious MCP tools and can scan tools, prompts, resources and server instructions.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Detection efficacy is not benchmarked publicly; the inspect API is a Cisco-hosted commercial dependency.", "capability": "multi-engine-scanning" }, { "id": "c2", "statement": "The repository states it scans MCP server source code for behavioural threats, detects mismatches between docstring claims and implementation with cross-file dataflow tracking, audits Python dependencies via pip-audit for CVE/PYSEC/GHSA issues, and hash-checks bundled binaries via VirusTotal.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "VirusTotal and pip-audit are external services; offline coverage and false-positive rates are not documented.", "capability": "code-and-dependency-analysis" }, { "id": "c3", "statement": "The GitHub repository metadata records an Apache-2.0 license, a non-archived state and commits in September 2026.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Repository activity does not establish maintenance commitments or release cadence.", "capability": "license-and-activity" } ], "limitations": [ "Best results depend on Cisco's commercial inspect API", "No published detection benchmarks" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": "Cisco", "integrations": [], "conflicts": [], "editorialNotes": "Distinct from Snyk Agent Scan: Cisco's tool emphasises code/dependency analysis rather than a runtime MCP proxy." }, { "name": "ClearML Model Registry", "id": "clearml-model-registry", "companyName": "ClearML", "companyUrl": "https://clear.ml/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "ML platform engineer", "MLOps lead", "Data science manager" ], "problemsSolved": [ "manual model tracking", "no provenance for deployed models", "automating promotion pipelines" ], "sources": [ { "id": "s1", "url": "https://clear.ml/docs/latest/docs/model_registry/", "title": "Model Registry — ClearML", "publisher": "ClearML", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents model cataloging with traceability and provenance, change documentation, lineage, project Model Table with customizable columns, querying, automatic/manual registration, CI/CD triggers on catalog changes, and Enterprise-plan launcher apps." }, { "id": "s2", "url": "https://github.com/clearml/clearml", "title": "clearml/clearml — MLOps/LLMOps solution", "publisher": "GitHub / ClearML", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Public repository describing experiment management, data management, pipelines, orchestration, scheduling and serving, model snapshot upload targets, provenance tracking and model monitoring with ClearML Serving." } ], "claims": [ { "id": "c1", "statement": "ClearML documentation states the model registry catalogs and shares models with traceability and provenance, tracks and documents model changes, shows model lineage, lists project models in a customizable Model Table with metadata and metric columns, supports querying by name, project, tag and metadata, supports automatic and manual registration, and can trigger CI/CD pipelines on catalog changes such as registering, tagging or publishing models.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs cover registry mechanics; approval workflows, risk assessment and regulatory mapping are not documented. Some launcher apps require the Enterprise plan.", "capability": "model-catalog-lineage-triggers" }, { "id": "c2", "statement": "The ClearML GitHub repository describes an open-source MLOps/LLMOps solution covering experiment management, data management, pipelines, orchestration, scheduling and serving, with optional automatic upload of model snapshots to shared folders, S3, GS, Azure or HTTP storage, model provenance tracking and out-of-the-box model monitoring via ClearML Serving.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Repository page confirms scope and storage targets; the licence text was not read, and enterprise-only features are not delineated there.", "capability": "open-source-toolchain" } ], "limitations": [ "Licence identifier not confirmed from the pages fetched this session.", "Split between open-source and Enterprise features is only partially documented on the pages read." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas", "library" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "productType 'hybrid' because docs evidence both open-source components and an Enterprise plan." }, { "name": "Collibra AI Command Center", "id": "collibra-ai-command-center", "companyName": "Collibra", "companyUrl": "https://www.collibra.com/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "model-governance", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief data officer", "Data governance lead", "AI governance lead" ], "problemsSolved": [ "disconnected AI and data governance", "no AI use-case register", "manual AI compliance assessments" ], "sources": [ { "id": "s1", "url": "https://productresources.collibra.com/docs/collibra/latest/Content/AIGovernance/co_about-ai-governance.htm", "title": "About Collibra AI Governance", "publisher": "Collibra", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes AI Governance community, AI Models and Agents domain with AI Model Version/AI Base Model/AI Project/AI Agent assets, AI Use Cases register, model and agent lifecycle stages, and states AI Governance has evolved into AI Command Center." }, { "id": "s2", "url": "https://www.collibra.com/products/ai-command-center", "title": "AI Command Center", "publisher": "Collibra", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims unified control plane for agents, models and use cases, agentic AI compliance assessment templates aligned to AI UC-1 available May 2026 with EU AI Act and NIST AI RMF templates, universal trust score and lifecycle tracking." }, { "id": "s3", "url": "https://www.collibra.com/company/newsroom/press-releases/collibra-introduces-collibra-ai-governance", "title": "Collibra Introduces Collibra AI Governance", "publisher": "Collibra", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Announces Collibra AI Governance built on the Collibra Data Intelligence Platform, with automated workflows, processes and policies to govern AI and its data; product shown at Gartner Data & Analytics Summit, March 2024." } ], "claims": [ { "id": "c1", "statement": "Collibra product documentation describes an out-of-the-box AI Governance community with an 'AI Models and Agents' technology asset domain (AI Model Version, AI Base Model, AI Project, AI Agent assets) and an 'AI Use Cases' business domain used as the AI use-case register, with lifecycle stages for models and agents.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation describes asset types and lifecycle stages; it does not establish automated discovery of models or agents.", "capability": "ai-asset-model-and-registry" }, { "id": "c2", "statement": "Product page states out-of-the-box agentic AI compliance assessment templates aligned to an emerging AI UC-1 standard available May 2026 alongside EU AI Act and NIST AI RMF templates, plus a universal trust score per system aggregating documentation, data integrity, lifecycle status and regulatory signals.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing page; template contents and score methodology are not published, and template availability dates mix announced and available features.", "capability": "assessment-templates-and-trust-score" }, { "id": "c3", "statement": "Documentation states that Collibra AI Governance has evolved into AI Command Center, which offers a more centralized experience; a 2024 press release documents the original AI Governance launch built on the Collibra Data Intelligence Platform.", "support": "documented", "sourceIds": [ "s1", "s3" ], "limitations": "Sources confirm the naming change but not the migration path or feature parity between the two.", "capability": "product-lineage-rename" } ], "limitations": [ "Some AI Command Center capabilities are dated to May 2026 availability on marketing pages without a release note confirming shipment.", "Requires the Collibra platform; standalone AI governance use is not evidenced." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [ "Collibra AI Governance" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Dedup: Collibra AI Governance and AI Command Center treated as one product with a rename, evidenced by Collibra docs." }, { "name": "Confident AI", "id": "confident-ai", "companyName": "Confident AI", "companyUrl": "https://www.confident-ai.com", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineering lead", "QA lead" ], "problemsSolved": [ "regression testing in CI", "dataset curation from traces", "agent trace inspection", "metric-based quality reporting" ], "sources": [ { "id": "s1", "url": "https://documentation.confident-ai.com/", "title": "Confident AI - The AI Quality Platform", "publisher": "Confident AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States every LLM call is captured as a trace with inputs, outputs, tool calls, latency, token cost and metadata; shows agent/tool/function spans with latency, token and cost totals; describes turning traces into datasets and running regression tests on pull requests." }, { "id": "s2", "url": "https://deepeval.com/docs/getting-started", "title": "DeepEval 5-min Quickstart", "publisher": "Confident AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Confident AI as an AI quality platform with observability, evals and monitoring, and DeepEval as the local evaluation runner that stores reports centrally in the cloud." } ], "claims": [ { "id": "c1", "statement": "Documentation states every LLM call is captured as a trace with inputs, outputs, tool calls, latency, token cost and metadata, with example traces showing agent, tool and function spans and totals for latency, tokens and cost.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Does not establish retention duration, tamper-evidence or export format for these records.", "capability": "trace-capture-detail" }, { "id": "c2", "statement": "The platform converts observability traces into evaluation datasets, auto-categorizes failures and edge cases, and runs regression tests on every pull request.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented policy gate that blocks deployment independently of the customer's own CI configuration.", "capability": "traces-to-datasets-and-ci" }, { "id": "c3", "statement": "Confident AI is presented as the AI quality platform built by the creators of the open-source DeepEval evaluation framework, which runs evaluations locally and can push reports to the cloud.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "The hosted platform itself is not documented as self-hostable.", "capability": "oss-core-relationship" } ], "limitations": [ "No documented audit logs, RBAC, SSO, retention configuration or PII redaction on the pages fetched.", "Buyer overlap with the DeepEval library entry; the two are recorded separately because one is a local library and one a hosted service." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "productType hybrid: commercial cloud platform with an open-source core (DeepEval) maintained by the same vendor." }, { "name": "ContextForge AI Gateway (mcp-context-forge)", "id": "ibm-contextforge-mcp-gateway", "companyName": "IBM", "companyUrl": "https://www.ibm.com/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "agent-discovery", "observability" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Platform engineering lead", "Enterprise architect", "DevOps engineer" ], "problemsSolved": [ "central MCP registry and proxy", "authentication and rate limiting for tool calls", "tool call observability" ], "sources": [ { "id": "s1", "url": "https://github.com/IBM/mcp-context-forge", "title": "IBM/mcp-context-forge: An AI Gateway, registry, and proxy", "publisher": "IBM (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Repository describes an open source registry and proxy federating MCP, A2A and REST/gRPC APIs with centralized governance, discovery and observability, plus plugin extensibility." }, { "id": "s2", "url": "https://ibm.github.io/mcp-context-forge/", "title": "ContextForge AI Gateway", "publisher": "IBM", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents Apache-2.0 licensing, PyPI/Docker/Helm distribution, Basic/JWT/custom auth, user-scoped OAuth tokens, tool input validation, rate limiting and OpenTelemetry instrumentation." } ], "claims": [ { "id": "c1", "statement": "ContextForge federates MCP servers, A2A servers and REST/gRPC APIs behind a unified endpoint with centralised discovery, governance and observability, and supports stdio, SSE and Streamable HTTP transports.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Repository documentation does not establish production scale limits or a formal support commitment.", "capability": "mcp-federation-proxy" }, { "id": "c2", "statement": "The gateway provides authentication (Basic, JWT or custom schemes), user-scoped OAuth tokens, rate limiting, retries and tool input validation with concurrency controls.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "It does not document semantic guardrails, prompt injection detection or human approval workflows.", "capability": "auth-and-rate-limits" }, { "id": "c3", "statement": "The project is distributed under the Apache License 2.0 and published as the PyPI package mcp-contextforge-gateway with OCI container images.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "License and packaging evidence does not indicate commercial support terms.", "capability": "apache-licensed" } ], "limitations": [ "No documented prompt-injection or content inspection capability", "Governance depends on how operators configure plugins" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Often referred to simply as 'MCP Gateway'; recorded under its current name ContextForge with the repository slug retained for identification." }, { "name": "Coval", "id": "coval", "companyName": "Coval", "companyUrl": "https://www.coval.dev", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Head of AI product", "Contact centre operations lead", "QA lead" ], "problemsSolved": [ "pre-launch simulation of conversational agents", "production call scoring", "regression detection across model and vendor changes", "human review routing" ], "sources": [ { "id": "s1", "url": "https://docs.coval.dev/", "title": "Welcome to Coval - Coval Documentation", "publisher": "Coval", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Coval as a deployment-readiness platform for voice AI agents, simulating thousands of realistic conversations with voice inputs, IVRs, edge cases, noise and accents, scoring every production call in real time, surfacing regressions in CI/CD, and routing failures to human reviewers." }, { "id": "s2", "url": "https://www.coval.dev/", "title": "Coval: Voice AI Testing & Evaluation Platform", "publisher": "Coval", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists simulation scenarios (accents, interruptions, policy traps, tool errors), production evals on live conversations, regression testing across prompt, model, vendor and workflow changes, evaluation dimensions including tool calls and caller emotion, and target users including compliance." } ], "claims": [ { "id": "c1", "statement": "Coval documents simulating thousands of realistic conversations before launch against voice or chat agents, including inbound, outbound and voice-to-voice connections, with edge cases, background noise, accents and IVRs.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs do not describe capture of internal agent spans or tool-call traces beyond conversation-level testing.", "capability": "conversation-simulation" }, { "id": "c2", "statement": "The platform scores every production call in real time, standardises metrics, surfaces regressions and can route failures to human reviewers whose feedback returns to the evaluation loop.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented retention period, immutable record or access-control model for stored call evaluations.", "capability": "production-scoring-and-review" }, { "id": "c3", "statement": "The product page states voice agent evaluation considers timing, turn-taking, interruptions, audio issues, tool calls, caller emotion and the final transcript, and that the same scenarios can be run across competing voice AI vendors.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing page; per-dimension scoring methodology is not documented.", "capability": "evaluation-dimensions" } ], "limitations": [ "No documented audit trail, RBAC, retention policy or PII redaction for call transcripts, which matters for regulated contact-centre use.", "Evidence is limited to voice/chat conversational agents; no coverage of back-office or coding agents." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Observability listed as secondary because production call scoring and QA review are documented alongside pre-launch simulation." }, { "name": "Credo AI Platform", "id": "credo-ai-platform", "companyName": "Credo AI", "companyUrl": "https://www.credo.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "agent-discovery", "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief AI officer", "AI governance lead", "Risk and compliance manager" ], "problemsSolved": [ "fragmented AI use-case intake", "manual control and evidence tracking", "third-party AI vendor review" ], "sources": [ { "id": "s1", "url": "https://knowledge.credo.ai/documentation", "title": "Documentation — Credo AI Platform", "publisher": "Credo AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists product documentation modules: reviews, task management, questionnaires, triggers and actions, governance status, governance plan, control library, risk scenario and control library, vendor registry (beta), roles and permissions." }, { "id": "s2", "url": "https://docs.sdk.credo.ai/", "title": "Introduction | Credo AI SDK Documentation", "publisher": "Credo AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes SDK capabilities: manage use cases, register models, track vendors, link models/vendors to use cases; Python and TypeScript clients; cookbooks for Jira and ServiceNow." }, { "id": "s3", "url": "https://www.credo.ai/", "title": "Credo AI — AI governance platform", "publisher": "Credo AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims AI/agent/vendor/model discovery and cataloging, agent registry and agent cards, shadow AI detection, risk classification, dependency mapping and policy packs for major regulations." } ], "claims": [ { "id": "c1", "statement": "Product documentation describes reviews, questionnaires, governance plans, governance status, a control library and a risk scenario library, plus triggers/actions and a beta vendor registry.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation index confirms these modules exist; it does not show coverage depth of the control library or any assessment of control effectiveness.", "capability": "governance-workflow-docs" }, { "id": "c2", "statement": "A documented SDK (Python and TypeScript) lets teams create and update use cases, register models and track vendors against the Credo AI API, with cookbooks referencing Jira and ServiceNow.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "SDK docs show API surface and example cookbooks; they do not establish supported production integrations or rate/scale limits.", "capability": "sdk-registry-api" }, { "id": "c3", "statement": "Vendor product page states the platform discovers and catalogs AI systems, agents, vendors and models, offers agent cards, shadow AI detection and prebuilt policy packs for frameworks including the EU AI Act and NIST AI RMF.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "Marketing page only. It does not evidence detection method, coverage, or that using policy packs makes an organisation EU AI Act compliant.", "capability": "agent-registry-policy-packs" } ], "limitations": [ "Vendor headquarters, pricing and deployment options were not stated on the pages fetched.", "No independent assessment of control-library quality or of governance outcomes was found." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Jira", "type": "documented_api", "sourceIds": [ "s2" ], "limitations": "Referenced as an SDK cookbook example, not evidenced as a supported native integration." }, { "name": "ServiceNow", "type": "documented_api", "sourceIds": [ "s2" ], "limitations": "Referenced as an SDK cookbook example only." }, { "name": "Databricks / MLflow", "type": "native", "sourceIds": [ "s1" ], "limitations": "Documentation index names a Databricks/MLflow integration article; the integration's scope was not read." } ], "conflicts": [], "editorialNotes": "Product family treated as one entry (Platform + SDK). SDK docs are a separate site (docs.sdk.credo.ai)." }, { "name": "CrewAI", "id": "crewai", "companyName": "CrewAI", "companyUrl": "https://www.crewai.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineer", "Automation lead", "Engineering manager" ], "problemsSolved": [ "multi-agent task orchestration", "deploying agents to managed infrastructure", "monitoring crew execution" ], "sources": [ { "id": "s1", "url": "https://docs.crewai.com/en/enterprise/introduction", "title": "CrewAI AMP", "publisher": "CrewAI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents CrewAI AMP (Agent Management Platform) for deploying, monitoring and scaling crews, with managed deployment via GitHub/Crew Studio/CLI, REST access, real-time event streaming, and detailed execution traces and logs." }, { "id": "s2", "url": "https://github.com/crewAIInc/crewAI", "title": "crewAIInc/crewAI: Framework for orchestrating role-playing, autonomous AI agents", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms the open-source framework for orchestrating role-playing autonomous agents, production-ready patterns with human input and checkpointing, task-level human review, and observability including metrics, logs and traces." } ], "claims": [ { "id": "c1", "statement": "The open-source framework orchestrates role-playing autonomous agents with production-ready patterns including deterministic steps, human input, structured outputs and checkpointing.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Repository documentation does not define enterprise policy controls around those patterns.", "capability": "role-based-agent-orchestration" }, { "id": "c2", "statement": "CrewAI AMP deploys crews to managed infrastructure via GitHub integration, Crew Studio or CLI, exposes them through generated REST endpoints, streams real-time events and provides detailed execution traces and logs.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The introduction page documents no audit-log, RBAC, versioning or tenant-isolation features.", "capability": "managed-deployment-and-traces" }, { "id": "c3", "statement": "Human-in-the-loop workflows are supported, with human review configurable on tasks.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Implemented in code per task; no reviewer role model or approval audit trail is documented.", "capability": "human-in-the-loop" }, { "id": "c4", "statement": "A CrewAI platform Helm chart with configuration and deployment guides exists for self-hosted installation.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "The self-hosted docs site was referenced in search results but its configuration details were not fetched in this session.", "capability": "self-hosted-platform" } ], "limitations": [ "Governance is developer-only in the sources reviewed: no approval console, RBAC, versioning or tenant-isolation documentation.", "Self-hosted platform configuration was not verified.", "Product naming shifted from CrewAI Enterprise to CrewAI AMP." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "saas", "self_hosted" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "CrewAI Enterprise" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "One entry for the OSS framework plus AMP platform; productType hybrid reflects the OSS core with a paid managed platform." }, { "name": "CrowdStrike AIDR Collectors", "id": "crowdstrike-aidr-collectors", "companyName": "CrowdStrike", "companyUrl": "https://www.crowdstrike.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "observability", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "SOC manager", "Endpoint security lead" ], "problemsSolved": [ "employee AI tool usage visibility", "AI traffic telemetry collection", "policy enforcement on AI data flows" ], "sources": [ { "id": "s1", "url": "https://aidr-docs.crowdstrike.com/docs/aidr/collectors/browser", "title": "AIDR - Browser Collectors", "publisher": "CrowdStrike", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the browser extension collector, supported browsers, MDM-based deployment, and the report-only limitation for output rules." }, { "id": "s2", "url": "https://aidr-docs.crowdstrike.com/docs/aidr/collectors/", "title": "AIDR - Collector Types", "publisher": "CrowdStrike", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists Falcon Endpoint, Browser, Application and Gateway collector categories including network inspection for desktop AI apps and named API gateways." } ], "claims": [ { "id": "c1", "statement": "A lightweight browser extension installed on enterprise-managed endpoints captures AI interactions in Chrome, Edge and Firefox and sends them to AIDR for visibility and policy enforcement.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation states browser collectors run output rules in report-only mode and do not block or redact model responses.", "capability": "browser-collector" }, { "id": "c2", "statement": "Collector categories documented are Falcon Endpoint (browser extension plus network inspection for desktop AI apps and coding assistants), Browser, Application via SDK/API, and Gateway for Apigee, Azure API Management, Kong and LiteLLM.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Docs do not state which AI providers are recognised by each collector or detection completeness.", "capability": "collector-types" }, { "id": "c3", "statement": "Extensions are deployed through enterprise management tooling such as Jamf, Microsoft Intune, Chrome Enterprise and Group Policy, with configuration applied through managed storage.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Unmanaged or BYOD endpoints are not covered by this deployment model.", "capability": "managed-deployment" } ], "limitations": [ "Discovery is of AI usage/traffic, not of an enterprise agent registry", "Response blocking is unavailable in browser collectors" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Kong AI Gateway", "type": "documented_api", "sourceIds": [ "s2" ], "limitations": "Listed as a supported gateway collector; depth of policy support not documented on this page." }, { "name": "Azure API Management", "type": "documented_api", "sourceIds": [ "s2" ], "limitations": "Listed as a supported gateway collector only." } ], "conflicts": [], "editorialNotes": "Entered as the AIDR collector layer because that is what the fetched documentation establishes; the broader AIDR product was not separately verified this session." }, { "name": "CSA STAR for AI", "id": "csa-star-for-ai", "companyName": "Cloud Security Alliance", "companyUrl": "https://cloudsecurityalliance.org", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "agent-security" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "CISO", "Head of Trust & Compliance", "AI product owner" ], "problemsSolved": [ "AI security transparency for customers", "controls self-assessment", "vendor assurance evidence" ], "sources": [ { "id": "s1", "url": "https://cloudsecurityalliance.org/star/ai", "title": "CSA STAR for AI", "publisher": "Cloud Security Alliance", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes STAR for AI levels, the AI-CAIQ self-assessment, Valid-AI-ted validation and the ISO/IEC 42001 reference for Level 2." }, { "id": "s2", "url": "https://cloudsecurityalliance.org/artifacts/ai-controls-matrix-v1-1", "title": "AI Controls Matrix v1.1", "publisher": "Cloud Security Alliance", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States 247 control objectives across 18 domains with mappings to ISO 42001, ISO 27001, BSI AIC4, the EU AI Act and NIST AI RMF." }, { "id": "s3", "url": "https://cloudsecurityalliance.org/press-releases/2026/03/10/csa-ai-controls-matrix-named-2026-cso-awards-winner", "title": "CSA AI Controls Matrix named 2026 CSO Awards winner", "publisher": "Cloud Security Alliance", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms the AICM is the foundation for STAR for AI and describes its coverage of the AI value chain." } ], "claims": [ { "id": "c1", "statement": "CSA describes STAR for AI with a Level 1 designation earned by completing and submitting the AI-CAIQ self-assessment to the STAR Registry, an enhanced Level 1 validated by the Valid-AI-ted automated system, and a Level 2 tier referencing ISO/IEC 42001 certification.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Level 2 requirements and auditor qualification rules were not fully verified; CSA is not an accreditation body for regulatory conformity assessment.", "capability": "program-levels" }, { "id": "c2", "statement": "CSA states the AI Controls Matrix contains 247 control objectives across 18 domains with mappings to ISO/IEC 42001, ISO/IEC 27001, BSI AIC4, the EU AI Act and NIST AI RMF, and that it is the foundation for STAR for AI.", "support": "documented", "sourceIds": [ "s2", "s3" ], "limitations": "The matrix is a control framework and mapping artefact; mappings are CSA's interpretation and are not endorsed by the mapped standards bodies.", "capability": "controls-foundation" } ], "limitations": [ "Level 1 is self-assessment, so submissions are not independently audited", "No accreditation-body oversight of the program evidenced" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "The AI Controls Matrix itself is a framework artefact, not a product, so it is recorded in the held file; the assurance program is the reviewable offering." }, { "name": "Cycode AI & ML Inventory and AIBOM", "id": "cycode-ai-ml-inventory", "companyName": "Cycode", "companyUrl": "https://cycode.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "ai-risk", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AppSec manager", "CISO", "Platform engineering lead" ], "problemsSolved": [ "shadow AI in the SDLC", "AI/ML asset inventory", "AIBOM for compliance" ], "sources": [ { "id": "s1", "url": "https://cycode.com/blog/tackle-shadow-ai-with-cycode-ai-ml-inventory/", "title": "Tackle Shadow AI with Cycode's AI & ML Inventory and AIBOM", "publisher": "Cycode", "type": "other", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the inventory pillars: discovery and mapping of AI infrastructure, models, coding assistants, packages and secrets across the SDLC; notes early access at publication." }, { "id": "s2", "url": "https://cycode.com/press/cycode-tackles-shadow-ai-with-launch-of-ai-inventory-and-aibom/", "title": "Cycode Tackles Shadow AI with Launch of AI Inventory and AIBOM", "publisher": "Cycode", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-10-21", "supports": "Press release announcing the AI/ML Inventory and AIBOM launch for discovering, governing and securing AI across the SDLC." } ], "claims": [ { "id": "c1", "statement": "Automatically discovers and catalogues AI and ML assets across the SDLC including AI infrastructure, models, coding assistants, packages and associated secrets.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Vendor blog; detection technique per asset class and supported ecosystems are not specified.", "capability": "sdlc-ai-discovery" }, { "id": "c2", "statement": "Cycode publicly launched AI and ML Inventory and AIBOM on 21 October 2025 as solutions for discovering, governing and securing AI use across the SDLC.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The launch blog describes the capability as being in early access at the time of writing; the press release announces launch without stating tier or region availability.", "capability": "launch-and-availability" } ], "limitations": [ "Documentation portal was not reachable during review, so claims rest on vendor blog and press release", "Early-access wording in the launch blog conflicts with press-release launch framing" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "IL", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "topic": "availability", "detail": "Vendor blog describes capabilities as in early access while the same-period press release announces launch; current GA status not confirmed from docs." } ], "editorialNotes": "Kept alongside Checkmarx because both scan the SDLC but Cycode's evidenced asset classes emphasise coding assistants and secrets rather than agents/MCP servers." }, { "name": "Cyera AI Guardian", "id": "cyera-ai-guardian", "companyName": "Cyera", "companyUrl": "https://www.cyera.com", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "agent-discovery", "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Cloud security architect", "Data security lead" ], "problemsSolved": [ "AI asset inventory", "shadow AI discovery", "sensitive data exposure to AI" ], "sources": [ { "id": "s1", "url": "https://www.cyera.com/latest-releases/ai-guardian", "title": "AI Guardian | Cyera Latest Releases", "publisher": "Cyera", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-08-12", "supports": "Announces AI Guardian as a unified solution combining AI-SPM and AI runtime protection across the AI lifecycle; gives no feature-level detail." }, { "id": "s2", "url": "https://www.cyera.com/platform/ai-spm", "title": "AI-SPM (AI Security Posture Management)", "publisher": "Cyera", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States AI-SPM is a capability within AI Guardian built on Cyera's DSPM foundation that discovers, classifies and inventories AI models, applications, agents and knowledge bases including shadow AI." } ], "claims": [ { "id": "c1", "statement": "AI-SPM automatically discovers, classifies and inventories AI assets including models, applications, agents, knowledge bases and copilots, covering sanctioned and shadow AI.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Product page only; no documentation of discovery connectors, refresh cadence or cloud coverage.", "capability": "ai-asset-inventory" }, { "id": "c2", "statement": "Sensitive data classifications for AI assets are supplied by Cyera's DSPM foundation, and AI-SPM is positioned as a capability within AI Guardian.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "The stated classification precision figure is vendor-reported and not independently verified.", "capability": "dspm-derived-classification" }, { "id": "c3", "statement": "AI Guardian is described as combining AI Security Posture Management with AI Runtime Protection across the AI lifecycle.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "The release page states no specific runtime controls, availability date or deployment model.", "capability": "ai-runtime-protection" } ], "limitations": [ "No public technical documentation was available, so runtime enforcement and policy mechanics are unverified.", "Availability and deployment model are not stated on either official page." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "IL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Cyera's generic DSPM was not counted separately; only the AI-specific AI Guardian/AI-SPM product is listed, per the AI-specific-evidence rule." }, { "name": "Databricks Unity Gateway", "id": "databricks-unity-gateway", "companyName": "Databricks", "companyUrl": "https://www.databricks.com/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "observability", "data-governance" ], "productType": "commercial", "availability": "unknown", "buyerRoles": [ "Data platform owner", "AI governance lead", "CISO" ], "problemsSolved": [ "identity-aware policy on model and tool calls", "MCP tool filtering", "AI spend and rate governance" ], "sources": [ { "id": "s1", "url": "https://www.databricks.com/product/artificial-intelligence/unity-gateway", "title": "Unity Gateway: Multi-AI governance and cost control", "publisher": "Databricks", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Unity Gateway as a centralized runtime governance layer for models, agents, MCP servers and AI tools with identity-aware policies, guardrails, budgets and audit visibility." }, { "id": "s2", "url": "https://docs.databricks.com/aws/en/ai-gateway/", "title": "AI governance with Unity Gateway", "publisher": "Databricks", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents governing models, model services and MCP services through Unity Catalog privileges, tool filtering, service policies evaluating request content and caller, rate limits and spend caps." } ], "claims": [ { "id": "c1", "statement": "Unity Gateway applies Unity Catalog privileges and service policies at request time to govern models, model provider services and MCP services, including tool filtering for MCP servers.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documentation does not state an availability stage (GA, beta or preview) for the gateway or its components.", "capability": "identity-aware-runtime-policy" }, { "id": "c2", "statement": "The product page states identity-aware policies control what agents can access, which actions they can take and which tools they can use, with agents inheriting user permissions on MCP connections.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No authorization API, token-exchange mechanism or decision format is specified.", "capability": "agent-action-controls" }, { "id": "c3", "statement": "The gateway can enforce rate limits, traffic splitting, fallbacks, per-user thresholds and hard spend caps for AI traffic.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Effectiveness for non-Databricks-hosted AI systems is asserted rather than documented.", "capability": "traffic-and-spend-limits" } ], "limitations": [ "Availability stage not stated in fetched documentation", "Coverage of third-party AI systems asserted on the product page only" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global", "US", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [ "Mosaic AI Gateway" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Databricks previously shipped 'AI Gateway' features in Mosaic AI Model Serving; the fetched docs path (/ai-gateway/) now documents Unity Gateway, recorded here as the current name. Also surfaced in agent-building research; runtime-controls kept as primary because the gateway's evidenced function is mediating and policing model/tool traffic." }, { "name": "Datadog Agent Observability", "id": "datadog-agent-observability", "companyName": "Datadog", "companyUrl": "https://www.datadoghq.com", "primaryCategoryId": "observability", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "SRE lead", "Platform engineering manager", "CISO" ], "problemsSolved": [ "structured span capture for agent runs", "token and cost telemetry", "incident investigation across AI and infrastructure telemetry" ], "sources": [ { "id": "s1", "url": "https://docs.datadoghq.com/llm_observability/", "title": "Agent Observability", "publisher": "Datadog", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Product index for Datadog Agent Observability with sections for Quickstart, Instrument, Investigate, Improve, Configure and Data Privacy, Security and RBAC." }, { "id": "s2", "url": "https://docs.datadoghq.com/llm_observability/setup/", "title": "Set up LLM Observability", "publisher": "Datadog", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents required span kinds (LLM, workflow, agent, tool, task, embedding, retrieval), rejection of invalid span kinds, detailed LLM span fields including tool calls, results and definitions, token and cost metric names, agent and tool span fields, and provider coverage." }, { "id": "s3", "url": "https://docs.datadoghq.com/llm_observability/instrumentation/", "title": "Agent Observability Instrumentation", "publisher": "Datadog", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Instrumentation index whose navigation exposes Audit Trail, Access Control, SSO with SAML, Data Security and Data Privacy, Security and RBAC entries." } ], "claims": [ { "id": "c1", "statement": "Setup documentation requires a span kind from LLM, workflow, agent, tool, task, embedding and retrieval, and states spans with an invalid span kind are not submitted.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Multi-agent handoff is not named as a first-class concept; it must be modelled with agent and workflow spans.", "capability": "span-taxonomy" }, { "id": "c2", "statement": "LLM spans capture invoked model name, operation name, model provider, user session ID, ML application name, input and output data, token metrics, cost metrics, metadata, tags, tool calls, tool results, tool definitions, prompt metadata and errors, with named token and cost metric fields; agent and tool spans capture function inputs, outputs, tags, metadata and errors.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Manually instrumented spans require the customer to record token counts, so completeness depends on instrumentation quality.", "capability": "llm-span-contents" }, { "id": "c3", "statement": "Product navigation includes a Data Privacy, Security and RBAC section, and platform administration documentation surfaces Audit Trail, Access Control, SSO with SAML and Data Security.", "support": "documented", "sourceIds": [ "s1", "s3" ], "limitations": "The pages fetched only expose navigation labels; retention periods, PII scrubbing behaviour and whether Audit Trail covers AI telemetry access are not established.", "capability": "platform-governance-surface" } ], "limitations": [ "Governance controls are inferred from navigation labels rather than feature documentation read in this session.", "Setup page carries a notice that the product is not supported for every Datadog site, so availability varies by region." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "LLM Observability" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Docs URL path remains /llm_observability/ while the product is titled Agent Observability; former name recorded." }, { "name": "Datawiza Agent Gateway", "id": "datawiza-agent-gateway", "companyName": "Datawiza", "companyUrl": "https://www.datawiza.com/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "identity-access" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "IAM manager", "Identity architect", "CISO" ], "problemsSolved": [ "IdP-backed authentication for MCP access", "tool discovery and invocation filtering", "least-privilege access to agent tools" ], "sources": [ { "id": "s1", "url": "https://www.datawiza.com/mcp-gateway", "title": "MCP Gateway: Identity-Native Security for MCP Servers", "publisher": "Datawiza", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Datawiza Agent Gateway as an identity-native MCP gateway validating IdP tokens and enforcing MCP server, tool and action policy on tools/list and tools/call." }, { "id": "s2", "url": "https://docs.datawiza.com/", "title": "Datawiza Documentation", "publisher": "Datawiza", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Datawiza as an identity-aware proxy and enterprise gateway for AI agents accessing tools, MCP and APIs, cloud native and cloud managed." } ], "claims": [ { "id": "c1", "statement": "The product page states users and agents sign in through the organisation's IdP before any request reaches an MCP server, with token issuer, audience, signature, expiry, scopes and claims validated.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Claim appears on a product page; no reference documentation for the validation pipeline was fetched.", "capability": "idp-gated-mcp-access" }, { "id": "c2", "statement": "Policy can filter which tools each identity discovers via tools/list and enforce which tools or actions it may invoke via tools/call, evaluating agent, claim, server, tool, action and environment.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Policy language, evaluation order and failure behaviour are not documented on the fetched pages.", "capability": "tool-list-and-call-policy" }, { "id": "c3", "statement": "Datawiza's documentation site describes the product as an identity-aware proxy and enterprise gateway for AI agents accessing tools, MCP and APIs, delivered cloud-native and cloud-managed.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The documentation landing page confirms scope but not specific enforcement features.", "capability": "agent-gateway-scope" } ], "limitations": [ "Enforcement detail rests largely on marketing material", "No evidence found on guardrails, content inspection or human approval" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Datawiza's older positioning was identity-aware proxy for web apps; the MCP gateway is described as a use case of the same Agent Gateway platform." }, { "name": "Daytona", "id": "daytona-sandboxes", "companyName": "Daytona", "companyUrl": "https://www.daytona.io/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "agent-building" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI platform engineer", "Head of engineering", "DevOps lead" ], "problemsSolved": [ "isolating AI-generated code execution", "resource and network limits for agent workloads", "stateful agent execution environments" ], "sources": [ { "id": "s1", "url": "https://www.daytona.io/", "title": "Daytona - Secure Infrastructure for Running AI-Generated Code", "publisher": "Daytona", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes secure elastic infrastructure for running AI-generated code with isolated runtime, fast sandbox creation, parallel execution and programmatic file, Git, LSP and execute APIs." }, { "id": "s2", "url": "https://github.com/daytonaio/daytona", "title": "daytonaio/daytona", "publisher": "Daytona (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States sandboxes provide complete isolation with dedicated kernel and network stack, lists API keys, limits, audit logs and network limits, and states the repository is no longer maintained as core development moved to a private codebase in June 2026." } ], "claims": [ { "id": "c1", "statement": "Daytona sandboxes are described as composable computers with complete isolation, a dedicated kernel, filesystem and network stack, and allocated vCPU, RAM and disk, created programmatically via SDKs.", "support": "documented", "sourceIds": [ "s2", "s1" ], "limitations": "Isolation claims are vendor-stated; no third-party security assessment was found.", "capability": "isolated-sandbox-execution" }, { "id": "c2", "statement": "Platform capabilities listed include API keys, limits, network limits, audit logs and webhooks around sandbox execution.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The controls are listed without documented policy semantics or enforcement examples.", "capability": "operational-controls" }, { "id": "c3", "statement": "The public repository states that as of June 2026 core development moved to a private codebase and the repository is no longer maintained, with no further updates, fixes or releases.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "This concerns the open-source distribution only; it does not indicate the commercial service is discontinued.", "capability": "open-source-status-changed" } ], "limitations": [ "Open-source repository unmaintained", "No tool-call authorization or threat detection capability evidenced" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "global", "US", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ "Marketing site presents an actively developed platform while the public repository declares itself unmaintained since June 2026" ], "editorialNotes": "Kept as verified because the commercial product pages and repository both describe the sandbox product; the open-source/commercial split is flagged in conflicts." }, { "name": "DeepEval", "id": "deepeval", "companyName": "Confident AI", "companyUrl": "https://www.confident-ai.com", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI engineer", "QA lead" ], "problemsSolved": [ "unit-style testing of LLM output", "agent trajectory and tool-call scoring", "RAG quality measurement", "regression testing in CI" ], "sources": [ { "id": "s1", "url": "https://github.com/confident-ai/deepeval", "title": "confident-ai/deepeval: The LLM Evaluation Framework", "publisher": "GitHub / Confident AI", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "README states end-to-end and component-level evaluation, trajectory evaluation across decisions and actions, capture of ordered model decisions, tool calls and intermediate steps, metric list, and framework integrations; states it is built by the founders of Confident AI." }, { "id": "s2", "url": "https://deepeval.com/docs/getting-started", "title": "DeepEval 5-min Quickstart", "publisher": "Confident AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States deepeval runs evaluations locally in the user environment, describes test-case fields, CLI login and API keys, and Confident AI as the cloud platform holding centralized testing reports." } ], "claims": [ { "id": "c1", "statement": "DeepEval supports end-to-end and component-level evaluation, scoring complete agent trajectories across decisions and actions as well as individual steps such as LLM calls, tool use, retrieval and sub-agent handoffs.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The repository README does not establish accuracy of the bundled metrics or any governance controls.", "capability": "trajectory-and-component-evaluation" }, { "id": "c2", "statement": "Evaluations run locally in the user's own environment, with an optional connection to the Confident AI cloud for centralized test reports.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Self-hosting of the cloud reporting component is not documented; no RBAC, SSO, retention or audit-log features are described.", "capability": "local-execution" }, { "id": "c3", "statement": "Documented evaluation integrations include OpenAI Agents, LangChain, LangGraph, CrewAI, Pydantic AI, LlamaIndex, Google ADK, AWS AgentCore and Mastra, plus MCP-related task-completion metrics.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Integration depth per framework is not described; presence in a list does not establish maintained parity.", "capability": "framework-coverage" } ], "limitations": [ "Developer testing library only: no audit trail, access control, retention policy or approval gating.", "License text was not read on the pages fetched, so the exact open-source license is unconfirmed." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "LangGraph", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Listed as an evaluation integration in the README; scope of instrumentation not detailed." }, { "name": "OpenAI Agents SDK", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Listed as an evaluation integration in the README; scope of instrumentation not detailed." } ], "conflicts": [], "editorialNotes": "Kept separate from the Confident AI platform entry: distinct artefact (local library, no account) with a different buyer than the hosted quality platform. Vendor relationship recorded in companyName." }, { "name": "Deeploy", "id": "deeploy", "companyName": "Deeploy", "companyUrl": "https://deeploy.ai/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "ai-risk", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "ML engineering lead", "Model risk manager", "AI governance lead" ], "problemsSolved": [ "ungated model deployment", "controls disconnected from deployment", "monitoring and logging of production AI" ], "sources": [ { "id": "s1", "url": "https://docs.deeploy.ai/", "title": "Embed governance in your AI deployments with Deeploy", "publisher": "Deeploy", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents model onboarding/registration, standardized risk assessments driving applicable controls, selectable global/European/company control frameworks, role-based deployment approval workflows, gateway for monitoring, guardrails, alerting, logging, authentication and load balancing, and supported artefact sources." }, { "id": "s2", "url": "https://deeploy.ai/", "title": "Deeploy — AI Governance Platform", "publisher": "Deeploy", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims runtime discovery, onboarding and monitoring of AI systems and agents, connection of any MLOps or GenAI platform, and centralised control without migration." } ], "claims": [ { "id": "c1", "statement": "Deeploy documentation states standardized assessments determine use-case risk levels, risk scores automatically determine which controls apply, control frameworks can be selected from global, European or company-specific sets, and approval workflows are configured by defining which roles must approve deployments in each workspace.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs describe configuration; they do not enumerate included framework content or prove that approvals are technically unbypassable.", "capability": "risk-scored-controls-and-approval-gates" }, { "id": "c2", "statement": "Documentation states model artefacts can be deployed from Git-based version control, MLflow, Databricks, Hugging Face or Azure model registry, models can be hosted as managed deployments, integrated as external APIs or registered, and an automated gateway provides monitoring, guardrailing, alerting, logging, authentication and load balancing.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Integration list is documented at a high level without version support details or limits.", "capability": "registry-integrations-and-gateway" }, { "id": "c3", "statement": "The vendor site claims discovery, onboarding and monitoring of any AI system or agent at runtime and connection of any MLOps or GenAI platform without migration.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing claim; 'any platform' breadth and discovery method are not evidenced.", "capability": "runtime-discovery-positioning" } ], "limitations": [ "Company headquarters and legal entity not stated on fetched pages (commonly reported as Netherlands).", "Self-hosted option inferred from runtime/gateway architecture references, not explicitly documented on pages read." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Docs exist on both docs.deeploy.ai and docs.deeploy.ml; the .ai domain was used." }, { "name": "Deloitte Algorithm Assurance", "id": "deloitte-algorithm-assurance", "companyName": "Deloitte", "companyUrl": "https://www.deloitte.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "model-governance", "evaluation" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Audit committee chair", "Chief Risk Officer", "Head of Analytics" ], "problemsSolved": [ "independent algorithm review", "bias and performance testing", "algorithm control frameworks" ], "sources": [ { "id": "s1", "url": "https://www.deloitte.com/au/en/services/audit-assurance/services/algorithm-assurance.html", "title": "Algorithm Assurance", "publisher": "Deloitte Australia", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Names Algorithm Assurance as a specialist assurance offering and lists its components and bias/regulatory assessment activities." }, { "id": "s2", "url": "https://www2.deloitte.com/ca/en/pages/audit/solutions/algorithmic-assurance.html", "title": "Algorithmic Assurance", "publisher": "Deloitte Canada", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes third-party code review, performance measurement and stress testing, process and control review, and logic/methodology assessment." } ], "claims": [ { "id": "c1", "statement": "Deloitte Australia describes Algorithm Assurance as a specialist assurance offering covering the definition, identification, classification, assessment, enhancement and monitoring of algorithms, including assessment of unintended bias and of compliance with regulatory requirements.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No assurance standard (for example ISAE 3000), sample report, or engagement duration is disclosed on the page.", "capability": "algorithm-assurance-scope" }, { "id": "c2", "statement": "Deloitte Canada's algorithmic assurance page states the service provides third-party review of the code underpinning an algorithm and measures algorithm performance under normal and stress-test conditions, plus process and control review.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Different member firm and practice (Actuarial, Rewards & Analytics); does not establish identical scope in other geographies.", "capability": "code-review-and-stress-testing" } ], "limitations": [ "Member-firm scope varies by country", "No named assurance standard or certification outcome" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "AU", "CA", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Australia and Canada pages describe the same named offering under slightly different labels (Algorithm Assurance / Algorithmic Assurance)." }, { "name": "Deloitte Trustworthy AI", "id": "deloitte-trustworthy-ai", "companyName": "Deloitte", "companyUrl": "https://www.deloitte.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "model-governance" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Chief Risk Officer", "Chief AI Officer", "Head of Internal Audit" ], "problemsSolved": [ "AI governance operating model design", "AI regulatory readiness", "model risk controls" ], "sources": [ { "id": "s1", "url": "https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/services/ethics-of-ai-framework.html", "title": "Trustworthy AI services", "publisher": "Deloitte", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists the Trustworthy AI framework and eight named offerings including AI Audit and Assurance and AI Model Risk Management." }, { "id": "s2", "url": "https://www.deloitte.com/us/en/services/consulting/services/ai-risk-governance-program.html", "title": "AI Risk and Governance Framework", "publisher": "Deloitte", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes use-case tiering, lifecycle checkpoints, application risk review, control design and testing, performance validation, and monitoring and response." } ], "claims": [ { "id": "c1", "statement": "Deloitte US lists Trustworthy AI offerings that include 'AI Risk Management & Governance', 'AI Regulatory Support Services', 'AI Model Risk Management' and 'AI Audit and Assurance'.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "The page names offerings without describing methodology, deliverables, staffing, or any independent assurance standard applied.", "capability": "named-offering-set" }, { "id": "c2", "statement": "The AI Risk Management and Governance service page describes use-case tiering by impact and data sensitivity, lifecycle gate reviews, control design and testing including human-in-the-loop and kill switch, and performance validation for accuracy and drift.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing service page; no sample deliverable, control catalogue, or third-party validation of the described controls.", "capability": "lifecycle-controls" } ], "limitations": [ "No published methodology or engagement scope", "No accreditation or attestation standard named for the 'AI Audit and Assurance' offering" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept separate from Deloitte Australia's Algorithm Assurance offering, which is a distinct named assurance service on a member-firm site. Deloitte trademarks 'Trustworthy AI'." }, { "name": "Descope Agentic Identity Hub", "id": "descope-agentic-identity-hub", "companyName": "Descope", "companyUrl": "https://www.descope.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Application security engineer", "Identity architect", "Platform engineer" ], "problemsSolved": [ "authenticating agents to APIs and MCP servers", "issuing short-lived agent credentials", "least-privilege policy for agent tool access" ], "sources": [ { "id": "s1", "url": "https://docs.descope.com/agentic-identity-hub", "title": "Agentic Identity Hub - Descope Documentation", "publisher": "Descope", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the Hub as Descope's control plane for AI agent identity covering agent clients, tools and credentials, with DCR/CIMD registration and runtime credential issuance." }, { "id": "s2", "url": "https://www.descope.com/press-release/agentic-identity-hub-2.5", "title": "Descope Unveils Agentic Identity Hub 2.5 With Enhanced Policy Capabilities", "publisher": "Descope", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-06-09", "supports": "Announces enhanced access policies, autonomous agent support and human-in-the-loop flows; dated Los Altos, CA, June 9, 2026." } ], "claims": [ { "id": "c1", "statement": "The Hub registers and manages OAuth clients for agents via pre-registration, DCR, CIMD or JWT bearer for cloud workloads, and creates agentic identity records on consent or tenant grant.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs do not state throughput limits or supported cloud providers beyond AWS and GCP OIDC.", "capability": "agent-client-registration" }, { "id": "c2", "statement": "Descope issues short-lived, scope-limited credentials at the moment an agent needs them.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Token lifetimes and revocation guarantees are not quantified in the fetched documentation.", "capability": "short-lived-scoped-credentials" }, { "id": "c3", "statement": "Version 2.5 added granular access policies for least-privilege agent access to MCP servers and backend APIs, using OAuth token exchange for delegated access with attribution and audits.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Press release announces capabilities without stating GA status for each one.", "capability": "agent-access-policies" } ], "limitations": [ "Deployment model beyond hosted SaaS was not stated on fetched pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Versioned releases 2.0 and 2.5 exist; treated as one product. Availability inferred from a dated press release describing shipped capabilities." }, { "name": "Dify", "id": "dify", "companyName": "Dify (LangGenius)", "companyUrl": "https://dify.ai", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineer", "Product engineer", "Technical founder" ], "problemsSolved": [ "building agentic workflows and RAG apps", "publishing AI apps as APIs or web apps", "monitoring LLM app performance" ], "sources": [ { "id": "s1", "url": "https://github.com/langgenius/dify", "title": "langgenius/dify: Build Agentic workflows, RAG pipelines", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms the repository for building agentic workflows and RAG pipelines, LLMOps monitoring of application logs and performance, and observability integrations with Opik, Langfuse and Arize Phoenix." }, { "id": "s2", "url": "https://docs.dify.ai/en/introduction", "title": "Introduction - Dify Docs", "publisher": "Dify", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Dify is an open-source platform for building AI applications including agents, agentic workflows and chatbots that use the user's own data, publishable as web apps or via REST API, with a difyctl CLI and plugin development." } ], "claims": [ { "id": "c1", "statement": "Dify is an open-source platform for building AI applications including agents, agentic workflows and chatbots that can draw on the user's own data and be published as web apps or integrated through a REST API, with a CLI for running apps from terminals, scripts, CI and AI agents.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documentation reviewed does not describe permission models or approval steps.", "capability": "agentic-workflow-building" }, { "id": "c2", "statement": "Provides observability features with integrations for Opik, Langfuse and Arize Phoenix, and LLMOps monitoring and analysis of application logs and performance over time.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Application logs are described for monitoring, not as a tamper-evident audit log.", "capability": "observability-integrations" }, { "id": "c3", "statement": "Human approval workflows, role-based access control, tenant isolation and audit logging are not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1", "s2" ], "limitations": "Absence of evidence; enterprise editions may document controls on pages not fetched.", "capability": "governance-controls" } ], "limitations": [ "No documented governance features: approvals, RBAC, versioning, tenant isolation and audit logs were all absent from the pages reviewed.", "The exact license terms (Dify uses a modified open-source license) were not verified in this session, so productType is recorded as hybrid." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Langfuse", "type": "native", "sourceIds": [ "s1" ], "limitations": "Listed as an observability integration in the repository README; configuration details were not fetched." } ], "conflicts": [], "editorialNotes": "Included for open-source and non-US coverage; governance gaps recorded explicitly in limitations per the category brief." }, { "name": "Dioptra", "id": "nist-dioptra", "companyName": "National Institute of Standards and Technology", "companyUrl": "https://www.nist.gov", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "ai-risk", "observability" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI evaluation lead", "Model risk manager", "Security researcher" ], "problemsSolved": [ "reproducible AI testing", "adversarial robustness experiments", "evidence tracking for AI risk measurement" ], "sources": [ { "id": "s1", "url": "https://github.com/usnistgov/dioptra", "title": "usnistgov/dioptra", "publisher": "NIST / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the platform, AI RMF Measure alignment, interfaces and reproducibility/traceability properties." }, { "id": "s2", "url": "https://pages.nist.gov/dioptra/", "title": "Dioptra documentation", "publisher": "NIST", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the microservice architecture and the measure/analyse/track AI risk purpose." } ], "claims": [ { "id": "c1", "statement": "The repository states Dioptra is a software test platform for assessing trustworthy characteristics of AI that supports the Measure function of the NIST AI Risk Management Framework and provides a REST API, web interface and Python client for designing, executing and tracking experiments.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "It measures characteristics for chosen experiments; it does not produce conformity statements or certifications.", "capability": "test-platform-scope" }, { "id": "c2", "statement": "NIST documentation describes Dioptra as a modular microservice environment for reproducible, trackable and reusable AI workflows to measure, analyse and track AI risks, with stated uses covering first-, second- and third-party model testing, research, evaluations and red-teaming.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documentation does not state the license name on the page and does not quantify supported model scale.", "capability": "reproducibility-and-use-cases" } ], "limitations": [ "Repository license metadata is NOASSERTION (US government work terms), so standard OSI licensing should not be assumed", "Operationally heavy: requires Docker-based deployment" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "License recorded cautiously: GitHub reports NOASSERTION and NIST directs users to license information shipped with the source." }, { "name": "DNV ISO/IEC 42001 certification and AI assurance", "id": "dnv-ai-assurance", "companyName": "DNV", "companyUrl": "https://www.dnv.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "model-governance" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Quality manager", "Procurement lead", "Head of Engineering assurance" ], "problemsSolved": [ "ISO/IEC 42001 certification", "AI vendor due diligence", "AI capability audit" ], "sources": [ { "id": "s1", "url": "https://www.dnv.co.uk/assurance/Management-Systems/42001-ai-management-campaign/42001-ai-management/", "title": "AI Governance and Assurance - ISO/IEC 42001", "publisher": "DNV", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes ISO/IEC 42001 third-party certification, an eight-step certification guide and self-assessment tool." }, { "id": "s2", "url": "https://www.dnv.com/digital-trust/services/ai-vendor-capability-assessment/", "title": "Independent AI vendor capability assessment", "publisher": "DNV", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the assessment as an independent third-party audit of AI/ML capability across governance, organisation and process topics." } ], "claims": [ { "id": "c1", "statement": "DNV describes third-party certification of an AI management system based on ISO/IEC 42001, supported by an eight-step guide and a self-assessment tool, and positions itself as a certification body for management systems.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No accreditation body is named for the ISO/IEC 42001 scheme on this page and certificate scope wording is not shown.", "capability": "aims-certification" }, { "id": "c2", "statement": "DNV states its AI vendor capability assessment is an independent third-party audit assessing capability to develop, deliver and operate trustworthy AI/ML and data-driven solutions across topics including governance, organisation and people, and processes.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The page says it covers 'eight key topics' while listing fewer; it is an assessment of organisational capability, not of a specific model's performance.", "capability": "vendor-capability-audit" } ], "limitations": [ "Accreditation for the AI scheme not evidenced", "Topic list on the vendor capability page is internally inconsistent (eight topics claimed, seven listed)" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "NO", "UK", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "issue": "Number of assessment topics", "detail": "DNV's page states the assessment covers 'the eight key topics of AI capabilities' but enumerates seven.", "sourceIds": [ "s2" ] } ], "editorialNotes": "Two distinct DNV offerings combined in one entry because both are AI assurance services from the same practice." }, { "name": "Docker MCP Gateway", "id": "docker-mcp-gateway", "companyName": "Docker, Inc.", "companyUrl": "https://www.docker.com/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "observability" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Developer platform lead", "Application security engineer", "DevOps engineer" ], "problemsSolved": [ "isolating MCP server execution", "central credential injection", "tool allowlisting for developer agents" ], "sources": [ { "id": "s1", "url": "https://docs.docker.com/ai/mcp-gateway/", "title": "MCP Gateway - Docker Docs", "publisher": "Docker, Inc.", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Docker's open source MCP Gateway as a centralized proxy managing server lifecycle, routing, authentication, credential injection, container isolation with restricted privileges and call tracing." }, { "id": "s2", "url": "https://github.com/docker/mcp-gateway", "title": "docker mcp CLI plugin / MCP Gateway", "publisher": "Docker (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms MIT licensing, per-server isolated containers, credential handling via Docker Desktop and profile-based tool allowlists with per-tool enable/disable." } ], "claims": [ { "id": "c1", "statement": "MCP servers run in isolated Docker containers with restricted privileges, network access and resource usage, with the gateway acting as a centralised proxy managing lifecycle, routing and authentication.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Documentation does not quantify isolation guarantees against container escape or cover remote third-party MCP servers.", "capability": "container-isolated-mcp-servers" }, { "id": "c2", "statement": "Tool allowlists are managed in profiles, and individual tools on a server can be enabled or disabled, determining which servers and tools clients can reach.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Allowlists are static configuration; no runtime policy evaluation on call arguments is documented.", "capability": "tool-allowlists" }, { "id": "c3", "statement": "The gateway injects required credentials before forwarding requests and provides logging and call-tracing for AI tool activity.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Log retention, export and SIEM integration are not documented on the fetched pages.", "capability": "credential-injection-and-tracing" } ], "limitations": [ "No prompt or content inspection", "No human approval workflow documented" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Shipped as the 'docker mcp' CLI plugin; recorded under the product name MCP Gateway used in Docker documentation." }, { "name": "Domino AI Governance", "id": "domino-ai-governance", "companyName": "Domino Data Lab", "companyUrl": "https://domino.ai/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Head of data science", "Model risk manager", "ML platform owner" ], "problemsSolved": [ "uncontrolled model promotion", "manual model documentation for reviews", "model lineage and audit trails" ], "sources": [ { "id": "s1", "url": "https://docs.dominodatalab.com/en/latest/user_guide/3b6ae5/manage-models-with-model-registry/", "title": "Manage models with model registry", "publisher": "Domino Data Lab", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents MLflow-based registry, project- and deployment-scoped discovery, metadata and lineage, custom model cards, version management, deployment to Domino-hosted or external endpoints, RBAC and project-role controls." }, { "id": "s2", "url": "https://docs.dominodatalab.com/en/latest/user_guide/1d334e/review-and-approve-models/", "title": "Review and approve models", "publisher": "Domino Data Lab", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents MLflow Model Registry in Domino as a centralized lifecycle hub with versioning and lineage, role-based permissions over stage transitions so only authorized personnel approve production moves, and a governance framework covering validation, ethical review, audit trails and stakeholder review." }, { "id": "s3", "url": "https://domino.ai/platform/ai-governance", "title": "AI governance platform | Domino Data Lab", "publisher": "Domino Data Lab", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims AI policies embedded in workflows and policy enforcement, central model registry, centralized policy management, self-documenting evidence, model cards for AI compliance, full model lineage and automated documentation." } ], "claims": [ { "id": "c1", "statement": "Domino documentation states its MLflow-based model registry supports project-scoped and deployment-scoped discovery, records model metadata and lineage, creates custom model cards, manages model versions, deploys to Domino-hosted or external endpoints, and uses RBAC and project roles to limit who can view, edit and collaborate on registered models.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs state RBAC over registry actions but do not document a formal multi-step approval workflow.", "capability": "mlflow-registry-with-rbac-and-model-cards" }, { "id": "c2", "statement": "Domino documentation on reviewing and approving models states role-based permissions regulate who can transition models into different stages so only authorized personnel can approve moves to production, and describes a governance framework including model validation, ethical review, audit trails and stakeholder review.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The page describes a governance framework and permission model, not automated enforcement against named regulations.", "capability": "role-based-stage-transitions-and-review" }, { "id": "c3", "statement": "Domino's product page claims AI policies embedded in workflows with policy enforcement, a central model registry, centralized policy management, self-documenting evidence, model cards for AI compliance, full model lineage and automated documentation.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "Marketing page; framework mappings and enforcement mechanics are not evidenced there.", "capability": "governance-product-positioning" } ], "limitations": [ "Registry capability is built on MLflow, so some behaviour is inherited from that project rather than Domino-specific.", "No evidence of EU AI Act / ISO 42001 control mappings on the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas", "hybrid" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Domino's earlier 'Domino Governance' / model governance docs and the current AI governance platform page treated as one product entry." }, { "name": "DutyGraph", "id": "dutygraph", "companyName": "DutyGraph", "companyUrl": "https://dutygraph.com", "primaryCategoryId": "work-delegation", "secondaryCategoryIds": [ "ai-risk" ], "productType": "commercial", "availability": "preview", "buyerRoles": [ "Advisor / consultant", "Business sponsor", "Operations lead" ], "problemsSolved": [ "Map how work actually happens before delegating it to AI", "Attach a named human owner and review checkpoint to a proposed agent task", "Surface handoffs and tasks with no agreed owner" ], "sources": [ { "id": "s1", "url": "https://dutygraph.com/landing/", "title": "DutyGraph — Make the work visible.", "publisher": "DutyGraph", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes advisor-led discovery across people, duties, tasks and software; task card contents; human/AI/AI-with-review labelling; an agent manifest with owner, task versions, requested access and review history; a founding pilot seeking 5-10 companies; and an explicit statement separating features available now from integration work not active in the pilot. Sample screens are labelled as a fictional company with actual app views." }, { "id": "s2", "url": "https://dutygraph.com/directory/ai-governance/", "title": "AI Governance Directory: Platforms, Roles & Evaluation Questions | DutyGraph", "publisher": "DutyGraph", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms DutyGraph publishes the directory, that the first edition lists 9 organizations reviewed September 6, 2026, that DutyGraph includes its own clearly labelled entry, that descriptions summarize sources rather than independent performance tests, and that the edition contains no paid placements." } ], "claims": [ { "id": "c1", "statement": "The product produces task cards recording the accountable owner, performer, required input, expected work product, software used, human instructions and a proposed AI prompt.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Evidenced only by the vendor's own landing page and a labelled sample screenshot; no documentation, API reference or independent verification was located.", "capability": "work-record-task-cards" }, { "id": "c2", "statement": "The vendor describes an agent manifest that carries the owner, task versions, requested access and review history, with a designated reviewer checking scope before authorization.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Described as a pilot capability alongside a fictional sample; the page does not evidence production use.", "capability": "delegation-boundary-review" }, { "id": "c3", "statement": "The vendor states that requests, versioned manifest drafts and a connected sample are available now, while live IAM provisioning, signed attestations and automatic revocation are integration work and not active features of the pilot.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "This is the vendor's own self-description of scope; it is useful as an explicit boundary statement but is not independent verification.", "capability": "shipped-vs-planned-scope" }, { "id": "c4", "statement": "DutyGraph publishes the editorial AI governance directory this research feeds, listing 9 organizations as of the September 6, 2026 review and stating there are no paid placements in the edition.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Confirms publisher role and edition size only; establishes nothing about DutyGraph's market position.", "capability": "directory-publisher" }, { "id": "c5", "statement": "Evidence of production deployments, named customers, funding, certifications or measured outcomes.", "support": "not_found", "sourceIds": [ "s1" ], "limitations": "The site states it is recruiting first pilot companies and that results and turnaround times are still being tested. 'not_found' means no supporting evidence was located, not that such evidence cannot exist.", "capability": "production-adoption" } ], "limitations": [ "All capability evidence is first-party marketing copy; no product documentation, API reference, changelog or trust center was located.", "Governance workflow is presented via a fictional sample company rather than a customer deployment.", "Named systems of record (Workday, Oracle HCM, Okta, Entra ID, Saviynt) appear only as sample records; the vendor does not claim live integrations and none should be inferred." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included per the brief because DutyGraph publishes the directory and self-lists. Held to the same evidence standard as every other entry: availability set to 'preview' on the strength of the vendor's own 'Advisor pilot' framing and its explicit not-active-features statement. Deliberately avoids describing DutyGraph as a leader, unique, or production-proven. The existing live directory uses a different, shorter category vocabulary (RUNTIME ACCESS, AI RISK & OVERSIGHT, IDENTITY & ACCESS, WORK DISCOVERY), so the 12-ID taxonomy in this research package requires a generator update before publication. The site's own maintenance note references 'Telarus supplier status', suggesting a channel relationship worth clarifying editorially before publishing partner-adjacent language." }, { "name": "E2B", "id": "e2b-sandboxes", "companyName": "E2B (e2b-dev)", "companyUrl": "https://e2b.dev/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "agent-building" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI platform engineer", "Head of engineering", "Application security engineer" ], "problemsSolved": [ "isolating agent code execution", "running untrusted generated code", "self-hosted execution boundary" ], "sources": [ { "id": "s1", "url": "https://docs.e2b.dev/", "title": "E2B Documentation", "publisher": "E2B", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States E2B provides isolated sandboxes that let agents safely execute code, process data and run tools, with SDKs to start and manage environments." }, { "id": "s2", "url": "https://github.com/e2b-dev/E2B", "title": "e2b-dev/E2B", "publisher": "E2B (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms Apache-2.0 licensing, open-source sandbox infrastructure for running AI-generated code, and Terraform-based self-hosting on AWS, GCP, Azure or Linux." } ], "claims": [ { "id": "c1", "statement": "E2B provides isolated sandboxes in which agents execute code, process data and run tools, started and managed through Python and JavaScript SDKs.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Documentation fetched does not detail the isolation technology or a formal security boundary assessment.", "capability": "isolated-agent-execution" }, { "id": "c2", "statement": "E2B infrastructure can be self-hosted using Terraform on AWS, Google Cloud, Azure or a general Linux machine, and the repository is Apache-2.0 licensed.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Self-hosted feature parity with the cloud service is not stated.", "capability": "self-hosting" } ], "limitations": [ "No tool-call authorization policy engine", "No agent threat detection capability evidenced" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included because execution boundaries/sandboxing is in scope for runtime controls; not a security-analytics product." }, { "name": "Entro Security NHI & Agentic AI Platform", "id": "entro-security", "companyName": "Entro Security", "companyUrl": "https://entro.security", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "agent-security" ], "productType": "commercial", "availability": "unknown", "buyerRoles": [ "CISO", "Security engineer", "Identity security lead" ], "problemsSolved": [ "agent and NHI inventory", "secret exposure and rotation", "detecting anomalous non-human behaviour" ], "sources": [ { "id": "s1", "url": "https://entro.security/platform-ai-agents/", "title": "Identify AI Agents. Secure Their NHIs.", "publisher": "Entro Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-03-07", "supports": "Describes agent inventory, shadow agent discovery, lineage mapping to NHIs, secrets, owners and creators, and NHIDR monitoring of agent activity." }, { "id": "s2", "url": "https://entro.security/", "title": "Entro Security: Agentic AI & Non-Human Identity Security Platform", "publisher": "Entro Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States discovery and inventory across clouds, code, CI/CD, on-prem and collaboration tools, ownership attribution, and NHIDR detection and response." } ], "claims": [ { "id": "c1", "statement": "Entro discovers and classifies agentic AI deployments, including shadow agents, and maps each agent to its non-human identities, entitlements, secrets, creator and human owner.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page does not name the agent platforms covered or the detection method per source.", "capability": "agentic-nhi-discovery" }, { "id": "c2", "statement": "The platform discovers NHIs and secrets across clouds, code, CI/CD, on-prem and collaboration tools and flags over-privileged, idle, misconfigured or orphaned tokens.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing site; no documentation of scanning coverage or false-positive handling.", "capability": "nhi-and-secret-inventory" }, { "id": "c3", "statement": "NHIDR continuously monitors agent permissions and activity and detects anomalous or risky behaviour of agents, NHIs and secrets in real time.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "No independent evaluation of detection efficacy was found.", "capability": "nhi-detection-response" } ], "limitations": [ "No official documentation site was fetched; both sources are marketing pages.", "No evidence of credential issuance or access-review workflows." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Vendor also markets 'Agentic Governance & Administration (AGA)'; not fetched, so not claimed here." }, { "name": "Enzai", "id": "enzai", "companyName": "Enzai", "companyUrl": "https://www.enz.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief risk officer", "AI governance lead", "Compliance manager" ], "problemsSolved": [ "undocumented AI governance decisions", "assessment tracking across many AI systems", "third-party AI oversight" ], "sources": [ { "id": "s1", "url": "https://www.enz.ai/", "title": "Enterprise AI Governance & Risk Management | Enzai", "publisher": "Enzai", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States centralized system of record for AI systems, models, datasets and governance decisions, support for EU AI Act, ISO 42001 and NIST RMF, third-party AI tracking, customizable intake with configurable fields and templates by system type and risk level." }, { "id": "s2", "url": "https://www.gov.uk/ai-assurance-techniques/enzai-ai-governance-hub", "title": "Enzai: AI Governance Hub", "publisher": "UK Government (DSIT)", "type": "other", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Case study describing the AI Governance Hub dashboard collating assessment results, per-system compliance status labels, real-time updates, performance monitoring against robustness/security/safety standards and selective sharing." }, { "id": "s3", "url": "https://www.enz.ai/product/agentic-ai-governance", "title": "Agentic AI Governance Platform", "publisher": "Enzai", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes autonomy risk tiering, action whitelisting at the boundary, escalation routing to humans, multi-agent visibility and recursion caps, and controls mapped to the OWASP Agentic Top 10." } ], "claims": [ { "id": "c1", "statement": "Homepage states a centralized system of record where AI systems, models, datasets and governance decisions are documented, assessed and auditable, with customizable intake using configurable fields and reusable templates tailored by system type and risk level, and support for EU AI Act, ISO 42001 and NIST RMF.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page. The claimed '50+ integrations' is not backed by a published integration list, and framework support does not establish compliance.", "capability": "system-of-record-and-intake" }, { "id": "c2", "statement": "A UK Government AI assurance techniques entry describes Enzai's AI Governance Hub collating results of every assessment run, recording each model or system as in full, partial or no compliance with regulation, updating in real time, and letting organisations choose with whom to share assessment information.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "gov.uk entry is a vendor-supplied case study; compliance labels are the vendor's own scoring, not an external audit.", "capability": "assessment-dashboard" }, { "id": "c3", "statement": "A vendor product page describes agentic AI governance features including autonomy risk tiering, action whitelisting, escalation routing, multi-agent visibility and controls mapped to the OWASP Agentic Top 10.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "Enforcement mechanism, integration points and whether these features are generally available are not evidenced; treat as marketing description.", "capability": "agentic-governance-module" } ], "limitations": [ "No technical documentation or API reference located; all claims are vendor-authored.", "Secondary runtime-controls categorisation rests on marketing description only." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Domain is enz.ai; company name Enzai. gov.uk entry names the product 'AI Governance Hub' while the current site markets the platform as 'Enzai'." }, { "name": "Evidently", "id": "evidently", "companyName": "Evidently AI", "companyUrl": "https://www.evidentlyai.com/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "observability", "evaluation" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "ML engineer", "Data scientist", "MLOps lead" ], "problemsSolved": [ "undetected data drift", "no regression tests for models in CI", "post-deployment quality monitoring" ], "sources": [ { "id": "s1", "url": "https://github.com/evidentlyai/evidently", "title": "evidentlyai/evidently — open-source ML and LLM observability framework", "publisher": "GitHub / Evidently AI", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents data drift detection with 20+ statistical tests and distance metrics, drift presets, test suites with pass/fail conditions for regression and CI/CD checks, monitoring UI over time; latest release v0.7.21." }, { "id": "s2", "url": "https://docs.evidentlyai.com/", "title": "Evidently documentation — introduction", "publisher": "Evidently AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States open source Apache 2.0, standalone Python library or self-hosted platform, 100+ metrics, declarative testing API, tracing, dataset management and dashboards for data and AI evaluation." } ], "claims": [ { "id": "c1", "statement": "The Evidently repository states support for data drift detection with 20+ statistical tests and distance metrics, drift presets such as PSI, and test suites with pass/fail conditions usable for regression testing, CI/CD checks and data validation, plus a monitoring UI visualising metrics and test results over time.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Repository documents metrics and tests; it documents no model registry, approvals or lifecycle governance, and no vendor legal entity is named on the page.", "capability": "drift-and-test-suites" }, { "id": "c2", "statement": "Evidently documentation states the product is open source under Apache 2.0, available as a standalone Python library or as part of a self-hosted platform, covering evaluation, testing and monitoring of data and AI systems with 100+ metrics, a declarative testing API, tracing and dashboards.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Docs confirm licence and deployment forms; managed cloud features and limits were not reviewed.", "capability": "license-and-deployment" } ], "limitations": [ "Categorised under model lifecycle governance for drift/performance monitoring; it does not provide registry, approval or regulatory features.", "Company details (legal entity, HQ) not stated on the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept as one entry covering the OSS library and self-hosted platform; Evidently Cloud not separately verified." }, { "name": "EY Responsible AI services", "id": "ey-responsible-ai-services", "companyName": "EY", "companyUrl": "https://www.ey.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Chief Risk Officer", "Chief AI Officer", "Head of Internal Audit" ], "problemsSolved": [ "AI readiness assessment", "GenAI governance framework design", "AI risk training" ], "sources": [ { "id": "s1", "url": "https://www.ey.com/en_uk/services/ai/responsible-ai", "title": "Responsible AI services", "publisher": "EY UK", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Names Responsible AI services, the Responsible AI Readiness Assessment with six readiness categories, the GenAI governance framework, and training programmes." }, { "id": "s2", "url": "https://www.ey.com/en_us/services/ai/responsible-ai", "title": "Responsible AI", "publisher": "EY US", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the EY Responsible AI framework as a systematic, technology-enabled approach spanning the AI lifecycle." } ], "claims": [ { "id": "c1", "statement": "EY UK states its Responsible AI Readiness Assessment gives a snapshot of readiness to manage AI risk and comply with forthcoming regulation, identifies gaps, and measures readiness across six categories.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "The six categories' criteria, scoring method and evidence requirements are not published.", "capability": "readiness-assessment" }, { "id": "c2", "statement": "EY describes a Responsible AI framework applied across design, development, deployment and monitoring, a GenAI governance framework for use-case prioritisation, and training and awareness programmes for stakeholders.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Marketing service pages only; no methodology, assurance standard, or independent review evidence.", "capability": "framework-and-training" } ], "limitations": [ "No published assessment criteria", "No evidence of independent assurance reporting" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "UK", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "EY also markets ey.ai platform assets; excluded here because this entry covers the advisory service, not software." }, { "name": "Fiddler", "id": "fiddler", "companyName": "Fiddler AI", "companyUrl": "https://www.fiddler.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Model risk manager", "ML platform lead", "AI governance lead" ], "problemsSolved": [ "LLM and ML production monitoring", "drift detection and root-cause analysis", "trust and safety metric enrichment", "pre- and post-deployment validation" ], "sources": [ { "id": "s1", "url": "https://docs.fiddler.ai/", "title": "Fiddler Documentation: Introduction to Fiddler", "publisher": "Fiddler AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the platform delivers visibility, context and control across traditional ML models, LLM applications and autonomous multi-agent systems, with monitoring of production agentic, GenAI and ML applications, experiments, evaluate and test, and guardrails sections." }, { "id": "s2", "url": "https://docs.fiddler.ai/product-guide/llm-monitoring", "title": "LLM Application Monitoring & Protection", "publisher": "Fiddler AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents publication of LLM inputs and outputs including prompts, prompt context, response and retrieved source documents; generated trust and safety metrics and enrichments; embedding enrichment, drift detection and UMAP embedding visualisations for root-cause analysis." } ], "claims": [ { "id": "c1", "statement": "Fiddler requires publication of LLM application inputs and outputs including prompts, prompt context, responses and, for RAG applications, retrieved source documents, and then generates trust and safety metrics and enrichments used for alerting, analysis and debugging.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The page does not document capture of LLM spans, tool calls or multi-agent handoffs, so agent-level traceability is unproven.", "capability": "llm-monitoring-inputs" }, { "id": "c2", "statement": "Documented capabilities include an embedding enrichment, drift detection and embedding visualisations with UMAP for root-cause analysis of problematic trends.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No documented retention, audit logging or access-control detail accompanying these analyses.", "capability": "drift-and-embeddings" }, { "id": "c3", "statement": "Product documentation describes monitoring traditional ML models, LLM applications and agentic applications in real time, experiments to test and validate LLM outputs, evaluation before and after deployment, and guardrails to protect AI applications.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Specific compliance certifications, RBAC design and policy-gating mechanics are not stated.", "capability": "scope-and-guardrails" } ], "limitations": [ "Model-monitoring lineage means metrics and drift are strong while agent trace structure (tool calls, handoffs) is not evidenced.", "No audit log, retention or certification evidence from the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept observability-primary with model-governance secondary; agent-specific claims deliberately not asserted since documentation covers prompts/responses rather than spans." }, { "name": "garak", "id": "garak", "companyName": "NVIDIA", "companyUrl": "https://www.nvidia.com", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "agent-security" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI security engineer", "Red team lead" ], "problemsSolved": [ "automated LLM vulnerability scanning", "jailbreak and prompt injection testing", "toxicity and data leakage probing" ], "sources": [ { "id": "s1", "url": "https://github.com/NVIDIA/garak", "title": "NVIDIA/garak: the LLM vulnerability scanner", "publisher": "GitHub / NVIDIA", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes garak as Generative AI Red-teaming & Assessment Kit under NVIDIA, its probe types and failure modes, report and hit-log outputs, default 10 generations per prompt, and supported model interfaces." }, { "id": "s2", "url": "https://docs.garak.ai/garak", "title": "Welcome to garak!", "publisher": "NVIDIA", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States garak discovers weaknesses and unwanted behaviours in systems using language model technology, scans a chatbot or model, produces a full report, and is Apache 2.0 licensed." } ], "claims": [ { "id": "c1", "statement": "garak combines static, dynamic and adaptive probes to test for hallucination, data leakage, prompt injection, misinformation, toxicity and jailbreaks, evaluates each probe with recommended detectors, and records generations both as received and when evaluated, with a separate hit log for successful attacks.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The README does not document tool-call or agent-trajectory testing, nor any coverage guarantee.", "capability": "probe-and-detector-scanning" }, { "id": "c2", "statement": "Supported targets include Hugging Face Hub models, OpenAI chat and completion APIs, AWS Bedrock foundation models, Replicate, LiteLLM, GGUF models via llama.cpp and arbitrary REST endpoints returning plaintext or JSON.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Support for a target does not establish probe effectiveness against it.", "capability": "target-coverage" }, { "id": "c3", "statement": "Documentation states garak is Apache 2.0 licensed and produces a full report detailing what worked and what needs improvement when scanning a chatbot or model.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The documentation site describes itself as a work in progress and states no governance or multi-user features.", "capability": "license-and-scope" } ], "limitations": [ "Command-line scanner for engineers: no multi-user access control, audit trail, retention policy or ticketing workflow.", "Documentation site links a legacy repository path (leondz/garak) while the maintained repository is NVIDIA/garak." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "description": "docs.garak.ai references the repository github.com/leondz/garak, whereas the maintained repository fetched is github.com/NVIDIA/garak.", "sourceIds": [ "s1", "s2" ] } ], "editorialNotes": "Project moved from the original author's namespace to NVIDIA; legacy link recorded as a conflict rather than a separate product. DEDUP FLAG: also present in services-oss.json. Recommend keeping one entry; evaluation is the correct primary category for a vulnerability scanner that tests models. Also surfaced in the open-source pass; evaluation kept as primary." }, { "name": "Gemini Enterprise Agent Platform", "id": "gemini-enterprise-agent-platform", "companyName": "Google Cloud", "companyUrl": "https://cloud.google.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "observability", "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI platform lead", "Cloud architect", "ML engineering manager" ], "problemsSolved": [ "deploying and scaling agents", "agent session and memory management", "agent evaluation and tracing" ], "sources": [ { "id": "s1", "url": "https://docs.cloud.google.com/agent-builder/agent-engine/overview", "title": "Scale your agents | Gemini Enterprise Agent Platform", "publisher": "Google Cloud", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the fully managed Agent Runtime with built-in observability, Sessions, Memory Bank, Example Store and Evaluation Service, framework support including ADK, LangGraph and A2A, and dynamic code execution." }, { "id": "s2", "url": "https://cloud.google.com/products/gemini-enterprise-agent-platform", "title": "Gemini Enterprise Agent Platform (formerly Vertex AI) | Google Cloud", "publisher": "Google Cloud", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms the rename from Vertex AI, positions the platform for building, scaling, governing and optimizing agents, and states custom-built agents can be registered, managed and governed via the Gemini Enterprise app." } ], "claims": [ { "id": "c1", "statement": "Provides a fully managed agent runtime to deploy and scale agents without managing infrastructure, with built-in observability and multiple deployment paths.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation does not state tenant isolation guarantees for the managed runtime.", "capability": "managed-agent-runtime" }, { "id": "c2", "statement": "Includes Sessions for continuous conversation state, a Memory Bank for persistent long-term memories with revision inspection, and an Example Store plus Evaluation Service to test, monitor and trace agent behaviour.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Memory retention policies and data residency controls are not described on this page.", "capability": "sessions-memory-and-evaluation" }, { "id": "c3", "statement": "Supports custom agents plus LangChain, LangGraph, LlamaIndex query pipeline, Agent2Agent, ADK and AG2 agents.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Feature parity across frameworks is not stated.", "capability": "multi-framework-support" }, { "id": "c4", "statement": "Custom-built agents can be securely registered, managed and governed using the Gemini Enterprise app.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Product page wording; the specific governance controls (permissions, approvals, audit) are not enumerated.", "capability": "agent-registration-and-governance" }, { "id": "c5", "statement": "A built-in human approval or action-gating workflow for agent tool calls is not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1", "s2" ], "limitations": "Absence of evidence; ADK may provide developer-level patterns not covered here.", "capability": "human-approval-workflow" } ], "limitations": [ "Documented governance is registration plus tracing; approvals, agent-level RBAC and audit trails were not evidenced in this session.", "Recent rebrand from Vertex AI means much third-party documentation refers to old product names." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "Vertex AI Agent Engine", "Vertex AI" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Single entry covering Agent Engine/Agent Platform Runtime and ADK, which the docs present as one stack; Gemini Enterprise (the end-user app) is a separate product not entered here." }, { "name": "Giskard Hub", "id": "giskard-hub", "companyName": "Giskard", "companyUrl": "https://www.giskard.ai", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "agent-security", "ai-risk" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI risk manager", "AI engineering lead", "CISO" ], "problemsSolved": [ "pre-deployment agent security testing", "hallucination and refusal testing", "RAG evaluation", "deployment sign-off evidence" ], "sources": [ { "id": "s1", "url": "https://docs.giskard.ai/en/latest/", "title": "Quickstart & setup | Giskard documentation", "publisher": "Giskard", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Giskard Open Source as a Python library for LLM testing and evaluation, LLM Scan with heuristic and LLM-assisted detectors, synthetic and adversarial test cases, and RAGET-generated question/reference answer/reference context test sets; notes Giskard Hub adds UI, collaboration and continuous red teaming." }, { "id": "s2", "url": "https://www.giskard.ai/", "title": "AI Red Teaming & LLM Security Platform | Giskard", "publisher": "Giskard", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Positions Giskard Hub as AI security delivered as a service with a severity-ranked vulnerability report, remediation guidance and a signed go/no-go deployment recommendation; lists tested failure categories." } ], "claims": [ { "id": "c1", "statement": "Giskard Open Source is a Python library whose LLM Scan combines heuristic and LLM-assisted detectors to generate synthetic and adversarial test cases for issues including prompt injection and stereotypes, while RAGET generates question, reference answer and reference context sets to test RAG agents.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The library documentation does not describe tracing, access control, retention or audit records.", "capability": "oss-scan-and-rag-testing" }, { "id": "c2", "statement": "The company positions Giskard Hub as AI security delivered as a service, producing a full vulnerability report ranked by severity, remediation guidance and a go/no-go deployment recommendation signed by its AI security experts.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Product page; the specific report contents, tester qualifications and cadence are not documented in reference material.", "capability": "assessment-and-report" }, { "id": "c3", "statement": "Documented test categories include sycophancy, prompt injection, data disclosure, inappropriate content, inappropriate denials, contradictions, hallucinations and omissions.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Category list appears on the marketing page without documented detection methodology per category.", "capability": "failure-taxonomy" } ], "limitations": [ "No documented audit logs, RBAC, retention or compliance certifications on the pages fetched.", "A dedicated red-teaming product page returned a client error, so Hub feature detail rests on the homepage." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "saas" ], "geographies": [ "EU", "FR", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "One entry covering the open-source library and the Hub service (hybrid). https://www.giskard.ai/products/ai-red-teaming returned HTTP client error on 2026-09-06." }, { "name": "GitGuardian NHI Governance", "id": "gitguardian-nhi-governance", "companyName": "GitGuardian", "companyUrl": "https://www.gitguardian.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Security engineer", "Platform engineer", "IAM manager" ], "problemsSolved": [ "secrets sprawl outside vaults", "stale and over-privileged machine credentials", "mapping NHI secrets to owners and consumers" ], "sources": [ { "id": "s1", "url": "https://www.gitguardian.com/nhi-governance", "title": "NHI Governance - NHI Security Solution for Enterprise", "publisher": "GitGuardian", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes inventory of NHI secrets, ownership mapping, stale/over-privileged detection and semi-automated rotation." }, { "id": "s2", "url": "https://docs.gitguardian.com/nhi-governance/integrate-your-sources", "title": "Integrate your NHI sources | GitGuardian documentation", "publisher": "GitGuardian", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-09-04", "supports": "Documents ggscout collection from secret managers, CI and infrastructure and permission/policy analysis of discovered credentials." } ], "claims": [ { "id": "c1", "statement": "NHI Governance uses the ggscout collector to inventory secrets and metadata from secrets managers, CI and infrastructure sources, and analyses the permissions and policies associated with discovered credentials.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documentation does not state discovery latency or completeness guarantees; no policy enforcement capability is documented.", "capability": "nhi-secret-inventory" }, { "id": "c2", "statement": "The product tracks consumers, rotation dates and permission levels, detects stale, unused and over-privileged secrets, and semi-automates rotation.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Rotation is described as semi-automated; supported target systems are not enumerated on the page.", "capability": "secret-lifecycle-management" } ], "limitations": [ "Agent-specific (as opposed to NHI/secret-specific) capabilities were not evidenced on fetched pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [ "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "French vendor; included for NHI credential governance, which is in scope, not for AI-agent identity issuance." }, { "name": "Google Cloud Agent Registry", "id": "google-agent-registry", "companyName": "Google", "companyUrl": "https://cloud.google.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "agent-building", "identity-access" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Platform engineering lead", "Cloud architect", "AI governance lead" ], "problemsSolved": [ "central agent and MCP server catalog", "agent discoverability across teams", "governed tool access for agents" ], "sources": [ { "id": "s1", "url": "https://docs.cloud.google.com/agent-registry/overview", "title": "Agent Registry overview", "publisher": "Google Cloud", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Defines Agent Registry as the governance pillar and unified inventory for agents, MCP servers, skills and endpoints, listing its API resource model." }, { "id": "s2", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/govern/agent-registry", "title": "Agent Registry | Gemini Enterprise Agent Platform", "publisher": "Google Cloud", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents registering agents automatically from supported runtimes or manually, registering MCP servers and endpoints, searching for agents and tools, and authenticating via auth manager and bindings." } ], "claims": [ { "id": "c1", "statement": "Agent Registry is a centralized catalog for storing, discovering and governing MCP servers, tools, standalone skills and AI agents in Google Cloud, exposing Agent, McpServer, Endpoint, Skill, SkillRevision and Publisher resources through its API.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation describes a registration-based model; it does not claim detection of unregistered or shadow agents.", "capability": "unified-agent-catalog" }, { "id": "c2", "statement": "Agents can be registered automatically from supported runtimes or manually for custom deployments, MCP servers and endpoints can be registered, and users can run keyword and prefix searches across registered capabilities.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Which runtimes support automatic registration is not enumerated on the pages reviewed.", "capability": "registration-and-search" }, { "id": "c3", "statement": "An auth manager and bindings let agents authenticate securely to discovered tools and resources.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Supported credential types and policy granularity are not described on this overview page.", "capability": "auth-bindings" } ], "limitations": [ "Inventory covers registered assets only", "Scoped to Google Cloud / Gemini Enterprise Agent Platform" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": "Alphabet", "integrations": [ { "name": "Agent Development Kit (ADK)", "type": "native", "sourceIds": [ "s2" ], "limitations": "Documented as registering custom ADK agents and resolving endpoints; other frameworks not evidenced here." } ], "conflicts": [], "editorialNotes": "Two Google documentation surfaces describe the same registry; treated as one product." }, { "name": "Google Cloud Model Armor", "id": "google-model-armor", "companyName": "Google", "companyUrl": "https://cloud.google.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "runtime-controls", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Cloud security architect", "AI platform owner", "CISO" ], "problemsSolved": [ "prompt injection and jailbreak screening", "sensitive data protection in prompts", "baseline safety policy across projects" ], "sources": [ { "id": "s1", "url": "https://docs.cloud.google.com/model-armor/overview", "title": "Model Armor overview", "publisher": "Google Cloud", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Model Armor as a Google Cloud service screening LLM prompts and responses, with prompt injection and jailbreak detection, sanitisation and blocking, usable for AI deployed in Google Cloud or other clouds." }, { "id": "s2", "url": "https://docs.cloud.google.com/model-armor/release-notes", "title": "Model Armor release notes", "publisher": "Google Cloud", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States GA for Google-managed MCP server and Gemini Enterprise Agent Platform integrations, preview for Agent Gateway integration, and lists prompt injection threat categories and floor settings for MCP traffic." } ], "claims": [ { "id": "c1", "statement": "Model Armor inspects incoming prompts and generated responses, can return sanitised versions, and blocks content when prompt injection or jailbreak detection is triggered.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The overview does not document tool-call authorization or agent action control.", "capability": "prompt-and-response-screening" }, { "id": "c2", "statement": "Release notes state integration with Google and Google Cloud MCP servers and with the Gemini Enterprise Agent Platform is generally available, floor settings define baseline filters for MCP server traffic, and Agent Gateway integration is in preview.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Preview features may change; coverage of third-party MCP servers is not established.", "capability": "mcp-and-agent-integrations" }, { "id": "c3", "statement": "Prompt injection and jailbreak detection flags threats including system instruction manipulation, unauthorized action execution and sensitive information retrieval.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Detection thresholds are configurable and effectiveness is not quantified.", "capability": "threat-categories" } ], "limitations": [ "No tool-call authorization or agent action blocking documented", "Agent Gateway integration in preview" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global", "US", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Model Armor is bundled with Security Command Center documentation; recorded as its own product entry because it is separately documented and callable." }, { "name": "GRACE Governance", "id": "grace-governance", "companyName": "2021.AI", "companyUrl": "https://2021.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief data officer", "Compliance officer", "AI programme manager" ], "problemsSolved": [ "undocumented AI approvals", "unclear AI ownership and accountability", "manual control reporting" ], "sources": [ { "id": "s1", "url": "https://2021.ai/products/grace-governance", "title": "GRACE Governance | Establish trust in AI", "publisher": "2021.AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists real-time monitoring, control management, AI registration and AI assessments; describes policy/control setup, AI registration, control reporting, compliance monitoring, audit trails and accountability roles." }, { "id": "s2", "url": "https://2021.ai/grace-ai-platform", "title": "GRACE AI Platform", "publisher": "2021.AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Presents GRACE modules including AI Registry (manage AI projects, risks, track AI operations) and AI Security (access, guardrails, AI detection, auto registration)." } ], "claims": [ { "id": "c1", "statement": "Product page states GRACE Governance covers policy and control definition, AI registration, AI assessments, control reporting, and monitoring and review of AI compliance, with roles, responsibilities and ownership defined across the governance process.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page with no technical documentation; assessment templates, framework coverage and evidence formats are not evidenced.", "capability": "policy-control-assessment-workflow" }, { "id": "c2", "statement": "Platform page positions GRACE Governance as one module of the GRACE AI Platform alongside an AI Registry for managing AI projects and risks and an AI Security area covering access, guardrails, AI detection and auto-registration.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Module boundaries and which capabilities require additional modules or licences are not stated.", "capability": "module-of-grace-platform" } ], "limitations": [ "No framework mapping (EU AI Act / ISO 42001) was evidenced on the pages fetched.", "No public documentation or release notes located." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Vendor is Danish (2021.AI); a partnership announcement with Velatir was seen in search results but not verified, so no integration recorded." }, { "name": "Harmonic Security", "id": "harmonic-security", "companyName": "Harmonic Security", "companyUrl": "https://www.harmonic.security", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "data-governance", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Security architect", "Data protection lead" ], "problemsSolved": [ "shadow AI usage visibility", "personal vs corporate AI account detection", "prompt data leakage prevention" ], "sources": [ { "id": "s1", "url": "https://www.harmonic.security/solutions/browser-based-genai-security", "title": "Browser-Based GenAI Security", "publisher": "Harmonic Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes browser-extension inventory of AI interactions, personal vs corporate account identification, embedded AI and AI browser detection, and nudge/justify/redirect controls." }, { "id": "s2", "url": "https://docs.harmonicsecurity.app/browser-extension", "title": "Frequently Asked Questions | Browser Extension", "publisher": "Harmonic Security", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents supported browsers and the exact telemetry captured: usage metadata rows and HTTP request payloads analysed for sensitive data." } ], "claims": [ { "id": "c1", "statement": "The browser extension monitors end-user web traffic and reports usage metadata (user, application, host, path, timestamp) plus filtered HTTP request payloads to identify which AI applications are in use.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Support documentation describes data captured; it does not quantify detection coverage of AI apps or handling of non-browser AI clients.", "capability": "browser-traffic-ai-inventory" }, { "id": "c2", "statement": "Identifies whether an AI tool session uses a personal or corporate account and which subscription plan is in use, and detects embedded AI features inside sanctioned SaaS apps and use of AI browsers.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Solution page claim; no documented method or accuracy data for plan-level detection.", "capability": "account-type-detection" }, { "id": "c3", "statement": "Extension support is documented for Chrome, Edge, Firefox, Safari, other Chromium browsers including Arc, Brave, Vivaldi, Island and Genspark, and AI browsers Comet and Dia, deployable via MDM tools such as Intune, Jamf and Kandji.", "support": "documented", "sourceIds": [ "s2", "s1" ], "limitations": "Feature parity across these browsers is not stated.", "capability": "browser-coverage" } ], "limitations": [ "Endpoint agents or network paths outside the managed browser are out of scope", "No evidence reviewed about discovering autonomous agents built on enterprise platforms" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "UK", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "UK-headquartered. Scope is employee AI tool usage, not enterprise agent inventory - recorded that boundary explicitly." }, { "name": "Helicone", "id": "helicone", "companyName": "Helicone", "companyUrl": "https://helicone.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineer", "Platform engineer" ], "problemsSolved": [ "logging LLM requests and responses", "multi-provider model routing", "usage and cost visibility" ], "sources": [ { "id": "s1", "url": "https://docs.helicone.ai/", "title": "Helicone quick start", "publisher": "Helicone", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the AI Gateway logging requests automatically, requests appearing in the Requests tab within seconds, access to 100+ models across providers, API key generation and an OpenAI-compatible chat completions endpoint." }, { "id": "s2", "url": "https://github.com/Helicone/helicone", "title": "Helicone/helicone", "publisher": "GitHub / Helicone", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Helicone is an open source LLM observability platform, Apache v2.0 licensed, one line of code to monitor, evaluate and experiment, SOC 2 and GDPR compliant, data ownership documentation and export via API or MCP server." } ], "claims": [ { "id": "c1", "statement": "Helicone's AI Gateway automatically logs LLM requests, which appear in the Requests tab within seconds, and provides access to 100+ models across providers including OpenAI, Anthropic, Google Vertex and Groq through an OpenAI-compatible endpoint.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The quickstart documents no LLM spans, tool calls, multi-agent handoffs, dataset scoring or evaluation features.", "capability": "request-logging" }, { "id": "c2", "statement": "The repository states Helicone is an open-source LLM observability platform licensed under Apache v2.0, references SOC 2 and GDPR compliance, and documents data export via API and access via an MCP server.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Compliance is asserted on the repository page without linked attestations; no RBAC, SSO, retention or PII redaction features are stated.", "capability": "license-and-compliance-claims" } ], "limitations": [ "Gateway-level logging captures requests, not agent decision structure, so multi-step agent accountability is limited.", "No access control, audit trail or retention documentation was found on the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "The docs introduction URL timed out twice; evidence taken from the docs quick-start page and the repository." }, { "name": "Holistic AI AI Audits", "id": "holistic-ai-audits", "companyName": "Holistic AI", "companyUrl": "https://www.holisticai.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "evaluation" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Head of Compliance", "General Counsel", "Chief AI Officer" ], "problemsSolved": [ "independent AI audit", "bias auditing for hiring laws", "DSA independent audit" ], "sources": [ { "id": "s1", "url": "https://www.holisticai.com/ai-audits", "title": "AI Audits - Mitigate & Monitor AI Risks", "publisher": "Holistic AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists audit verticals (bias, privacy, efficacy, robustness, explainability), regulation-specific assessments and reporting deliverables." }, { "id": "s2", "url": "https://www.holisticai.com/digital-services-act-audit", "title": "Independent Audit under the Digital Services Act", "publisher": "Holistic AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes independent DSA audits for VLOPs/VLOSEs and cites the Wikimedia Foundation engagement." } ], "claims": [ { "id": "c1", "statement": "Holistic AI states its AI audits evaluate systems for bias, privacy, efficacy, robustness and explainability, include regulation-specific assessments, and deliver audit reporting with context-specific impact analysis and custom mitigation strategies.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Service page only; audit standard, sampling method, assurance level and auditor independence controls are not documented.", "capability": "audit-scope" }, { "id": "c2", "statement": "Holistic AI offers independent audits of Digital Services Act due-diligence obligations for intermediary services, VLOPs and VLOSEs, and states it conducted the audit for the Wikimedia Foundation described as the first independent audit under the DSA.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Vendor page describes the engagement; the audit report itself and any regulator acceptance were not verified here.", "capability": "dsa-independent-audit" } ], "limitations": [ "No accreditation or professional-body oversight evidenced", "Holistic AI also sells a governance platform, so audit independence boundaries are not documented publicly" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "UK", "US", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Only the audit/assurance services are in scope here; Holistic AI's governance software belongs in a software category and is excluded from this entry." }, { "name": "Holistic AI Governance Platform", "id": "holistic-ai-governance-platform", "companyName": "Holistic AI", "companyUrl": "https://www.holisticai.com/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "agent-discovery", "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief risk officer", "AI governance lead", "Procurement / third-party risk" ], "problemsSolved": [ "ungoverned or shadow AI systems", "inconsistent AI risk scoring", "third-party AI procurement review" ], "sources": [ { "id": "s1", "url": "https://www.holisticai.com/ai-governance-platform", "title": "An End to End AI Governance Platform", "publisher": "Holistic AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes identify/inventory stage: shadow AI discovery across cloud, code and SaaS, centralised inventory of models, agents, datasets and endpoints, classification and lifecycle tracking, 15+ integrations, read-only." }, { "id": "s2", "url": "https://www.gov.uk/ai-assurance-techniques/holistic-ai-governance-risk-and-compliance-platform", "title": "Holistic AI: Governance, Risk and Compliance Platform", "publisher": "UK Government (DSIT)", "type": "other", "accessedAt": "2026-09-06", "publishedAt": "2023-09-19", "supports": "Case-study entry describing SaaS platform, low/medium/high risk levels, RAG dashboard, separate risk verticals for bias, robustness, efficacy, transparency and privacy, and third-party risk use." } ], "claims": [ { "id": "c1", "statement": "Product page states the platform continuously discovers and inventories models, agents, APIs and pipelines across cloud, code and vendors, including shadow AI, with read-only, agentless scanning.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page; scan coverage, supported clouds and detection accuracy are not evidenced, and 'no agents to install' is unverified.", "capability": "ai-inventory-discovery" }, { "id": "c2", "statement": "A UK Government AI assurance techniques entry describes the platform assigning low/medium/high risk levels shown on a single red-amber-green dashboard, with separate ratings for bias, robustness, efficacy, transparency and privacy, and assessment of safeguards.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "The gov.uk entry is a vendor-submitted case study published in 2023 by DSIT; it is not an independent audit and may not reflect the current product.", "capability": "risk-verticals-scoring" } ], "limitations": [ "Integration count (15+) is asserted by the vendor without a documented integration list on the pages fetched.", "gov.uk listing is from 2023 and may lag current capability." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Vendor markets itself as UK-based, but headquarters were not stated on the pages fetched; geography left empty." }, { "name": "HoneyHive", "id": "honeyhive", "companyName": "HoneyHive", "companyUrl": "https://www.honeyhive.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI platform lead", "Risk and governance manager", "AI engineer" ], "problemsSolved": [ "step-by-step replay of agent runs", "monitoring agents inside third-party enterprise platforms", "expert annotation to datasets", "online evaluation of live traffic" ], "sources": [ { "id": "s1", "url": "https://docs.honeyhive.ai/introduction", "title": "What is HoneyHive?", "publisher": "HoneyHive", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes HoneyHive as an AI observability and evaluation platform with distributed tracing capturing every interaction and step, agent graphs and threads, curated datasets from failing traces, experiments tracking regressions, annotation queues and online evals on production traces." }, { "id": "s2", "url": "https://www.honeyhive.ai/", "title": "HoneyHive AI", "publisher": "HoneyHive", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States agent observability for mission-critical AI, step-by-step replay of every tool call, prompt and decision, trajectories spanning hours or days, retries, loops and sub-agent handoffs, observation of agents inside ServiceNow, Microsoft Copilot and Salesforce, and named enterprise usage with scale figures." } ], "claims": [ { "id": "c1", "statement": "HoneyHive documents distributed tracing that captures every interaction and every step of an AI application, with agent graphs and threads, curated datasets from failing production traces, experiments to track regressions, annotation queues for expert feedback, and online evals on production traces.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The docs introduction does not itemise span kinds, cost or token telemetry, retention or access controls.", "capability": "distributed-tracing" }, { "id": "c2", "statement": "The product page states runs can be replayed step by step showing every tool call, prompt and decision in order, trajectories can span hours or days, and captured behaviour includes retries, loops and handoffs between sub-agents, including agents inside ServiceNow, Microsoft Copilot and Salesforce.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing page; instrumentation method for third-party platforms and the fidelity of replay are not documented in reference material.", "capability": "replay-and-handoffs" }, { "id": "c3", "statement": "The product page states HoneyHive powers observability and evaluation across dozens of mission-critical AI applications at Commonwealth Bank of Australia, serving agents used by 17M retail consumers and 55K internal users.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Customer scale figures are vendor-stated and not independently corroborated.", "capability": "enterprise-usage" } ], "limitations": [ "No audit log, retention, RBAC or PII-redaction evidence: the security documentation page could not be retrieved.", "Governance framing is addressed at the buyer level (risk and governance teams named) rather than through documented controls." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "AU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "docs.honeyhive.ai/security and docs.honeyhive.ai/observability/introduction both timed out on 2026-09-06; unreachable source retained for transparency and not cited in support of any capability." }, { "name": "IAPP Artificial Intelligence Governance Professional (AIGP)", "id": "iapp-aigp", "companyName": "IAPP", "companyUrl": "https://iapp.org", "primaryCategoryId": "services", "secondaryCategoryIds": [], "productType": "service", "availability": "generally_available", "buyerRoles": [ "AI governance lead", "Privacy officer", "Compliance manager" ], "problemsSolved": [ "AI governance skills certification", "team upskilling on AI law", "hiring signal for governance roles" ], "sources": [ { "id": "s1", "url": "https://iapp.org/certify/aigp", "title": "AIGP: Artificial Intelligence Governance Professional", "publisher": "IAPP", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the credential, certification process, body of knowledge and exam blueprint and study resources." }, { "id": "s2", "url": "https://store.iapp.org/ai-governance-professional-aigp-online-training/", "title": "AI Governance Professional (AIGP) Online Training", "publisher": "IAPP", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the training is aligned with the AIGP certification program and lists learning objectives." } ], "claims": [ { "id": "c1", "statement": "IAPP states the AIGP credential demonstrates competency in AI development, ethical AI deployment and AI management best practices, supported by a published AIGP Body of Knowledge and Exam Blueprint, study guides and self-assessment tests.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No accreditation body is named for the certification scheme, and the page states no geographic scope.", "capability": "credential-scope" }, { "id": "c2", "statement": "IAPP sells AIGP online training aligned with the certification program, aimed at professionals implementing AI governance and risk management, covering AI foundations, the development lifecycle, applicable laws and responsible governance implementation.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Training is IAPP's own; it does not evidence exam pass rates, market recognition, or regulator endorsement.", "capability": "aligned-training" } ], "limitations": [ "No external accreditation of the certification scheme evidenced", "Credential attests to individual knowledge only" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "US", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "ISACA's Advanced in AI Audit (AAIA) credential was verified but held to avoid over-representing individual credentials in this directory." }, { "name": "IBM watsonx.governance", "id": "ibm-watsonx-governance", "companyName": "IBM", "companyUrl": "https://www.ibm.com/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "model-governance", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief risk officer", "Model risk manager", "AI platform owner" ], "problemsSolved": [ "untracked models and prompt templates", "undetected drift and fairness degradation", "audit-ready AI documentation" ], "sources": [ { "id": "s1", "url": "https://www.ibm.com/docs/en/watsonx/saas?topic=governing-ai", "title": "Governing assets with watsonx.governance", "publisher": "IBM", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes end-to-end monitoring for ML and generative AI, OpenScale monitor configuration and thresholds, fairness and drift alerts, model health monitor, foundation-model alerts for abusive language and PII, AI Factsheets metadata capture and a common model inventory dashboard." }, { "id": "s2", "url": "https://www.ibm.com/products/watsonx-governance", "title": "IBM watsonx.governance", "publisher": "IBM", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims governance graph, shadow AI detection, broad regulatory/risk content library and enterprise GRC integration; cites 2026 Gartner Magic Quadrant leader placement." } ], "claims": [ { "id": "c1", "statement": "IBM documentation states AI Factsheets collects metadata for machine learning models and explicitly tracked prompt templates, and that a common model inventory dashboard shows all tracked assets.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs describe metadata capture and inventory views for assets tracked in watsonx; coverage of AI running outside IBM environments is not established here.", "capability": "factsheets-and-model-inventory" }, { "id": "c2", "statement": "IBM documentation states end-to-end monitoring for machine learning and generative AI, with configurable monitors and thresholds for fairness and accuracy drift, a model health monitor for real-time performance, and foundation-model alerts for hateful/abusive language or detected personal information.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation describes monitor availability and alerting, not detection accuracy or benchmark results.", "capability": "monitors-and-alerts" }, { "id": "c3", "statement": "Product page claims a governance graph visualising relationships between AI systems, risks, controls and policies, plus shadow AI detection of unapproved AI usage, and cites Gartner Magic Quadrant leader placement for AI governance platforms in 2026.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing page; discovery method and coverage are unspecified, and analyst placement is not evidence of a technical capability.", "capability": "governance-graph-and-shadow-ai" } ], "limitations": [ "Vendor-stated efficiency and audit-fee reduction figures were unsourced placeholders on the product page and are excluded.", "Self-hosted availability is inferred from the existence of Software Hub docs; the fetched pages covered SaaS." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Watson OpenScale functionality is documented inside watsonx.governance; treated as one product entry." }, { "name": "Idira Secure AI Agents", "id": "idira-secure-ai-agents", "companyName": "Palo Alto Networks", "companyUrl": "https://www.paloaltonetworks.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "PAM administrator", "IAM manager" ], "problemsSolved": [ "discovering active AI agents", "eliminating standing privilege for agents", "auditing agent actions and on-behalf-of context" ], "sources": [ { "id": "s1", "url": "https://www.paloaltonetworks.com/idira/agentic", "title": "Agentic Identity Security - Palo Alto Networks", "publisher": "Palo Alto Networks", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes agent scanning and enrichment, the agent identity broker, task-duration access, automatic revocation and logging of agent actions." }, { "id": "s2", "url": "https://siliconangle.com/2026/05/12/idira-launches-palo-alto-networks-extends-cyberark-tech-machine-agentic-identities/", "title": "Idira launches as Palo Alto Networks extends CyberArk tech to machine and agentic identities", "publisher": "SiliconANGLE", "type": "independent_reporting", "accessedAt": "2026-09-06", "publishedAt": "2026-05-12", "supports": "Reports Idira is generally available, unifies human, machine and AI agent identities and applies zero standing privilege and just-in-time access." } ], "claims": [ { "id": "c1", "statement": "Idira Secure AI Agents scans SaaS, cloud and developer environments to identify active agents and enriches them with ownership, purpose, status and permission level.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page does not name scanned platforms or detection methods.", "capability": "agent-discovery-with-context" }, { "id": "c2", "statement": "A dedicated agent identity broker acts as a dynamic enforcement point granting agents access only for the duration of a task and revoking permissions when the job is complete.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Supported target systems for brokered access are not listed.", "capability": "agent-identity-broker" }, { "id": "c3", "statement": "Idira launched as a generally available identity security platform extending CyberArk privileged access controls to machine and AI agent identities, with further features slated for later in the year.", "support": "independently_supported", "sourceIds": [ "s2" ], "limitations": "Trade press report; feature-level GA status per module is not broken out.", "capability": "platform-availability" } ], "limitations": [ "CyberArk-branded agent pages now redirect to Idira; older CyberArk documentation could not be verified as current." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [ "CyberArk Secure AI Agents" ], "parentCompany": "Palo Alto Networks", "integrations": [], "conflicts": [], "editorialNotes": "Dedup/acquisition decision: CyberArk 'Secure AI Agents' pages redirect to Palo Alto Networks Idira, so this replaces a separate CyberArk entry." }, { "name": "Immuta", "id": "immuta", "companyName": "Immuta", "companyUrl": "https://www.immuta.com", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "identity-access" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Data governance lead", "Data platform owner", "CISO" ], "problemsSolved": [ "data access policy enforcement", "classification and metadata management", "access monitoring and audit" ], "sources": [ { "id": "s1", "url": "https://www.immuta.com/product/data-security-ai/", "title": "Data Security for AI - Immuta", "publisher": "Immuta", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2024-11-01", "supports": "States RAG indexes and storage platforms can be onboarded as data sources, and describes classification, data-layer access controls enforced at query time, real-time monitoring and auditing." }, { "id": "s2", "url": "https://documentation.immuta.com/SaaS/", "title": "Immuta Documentation - SaaS", "publisher": "Immuta", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms the SaaS product documentation covers governed data access with built-in controls, access request management and unified data access control across platforms." } ], "claims": [ { "id": "c1", "statement": "RAG indexes and storage platforms can be onboarded and integrated as new Immuta data sources.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page states the capability without naming supported vector stores or describing how policies apply to embeddings.", "capability": "rag-index-onboarding" }, { "id": "c2", "statement": "Access controls are defined at the data layer and enforced when data is queried, with row-level classifications and domain-scoped policies across platforms such as Snowflake and Databricks.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Not an AI-specific control; the page does not cover enforcement for non-SQL AI retrieval paths.", "capability": "query-time-policy-enforcement" }, { "id": "c3", "statement": "Monitors data access queries in real time, alerts on unusual behaviour, and provides a single view of activity for auditing usage.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Audit retention and export mechanisms are not described.", "capability": "access-monitoring-and-audit" }, { "id": "c4", "statement": "Documentation positions Immuta as providing governed access to data with built-in controls, collaboration workflows and automated provisioning across data platforms.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The documentation landing page contains no AI- or agent-specific governance content.", "capability": "governed-access-provisioning" } ], "limitations": [ "AI-specific evidence is thin: only the RAG index onboarding statement, on a 2024 product page.", "No documentation of prompt-level redaction, agent identity handling or AI lineage." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Snowflake", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Named as a platform where domain policies apply; no integration reference documentation was fetched." }, { "name": "Databricks", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Named on the product page only." } ], "conflicts": [], "editorialNotes": "Borderline under the AI-specific-evidence rule; retained because the official product page explicitly names RAG index onboarding. Immuta's separately announced 'Immuta AI' copilot features were not verified and are noted in the held list." }, { "name": "Infosys Topaz Responsible AI Suite", "id": "infosys-responsible-ai-suite", "companyName": "Infosys", "companyUrl": "https://www.infosys.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "observability" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Chief Data Officer", "Chief Risk Officer", "Head of AI platform" ], "problemsSolved": [ "responsible AI implementation", "AI compliance monitoring", "AI risk assessment" ], "sources": [ { "id": "s1", "url": "https://www.infosys.com/services/data-ai-topaz/offerings/responsible-ai-overview.html", "title": "Responsible AI Solutions & Framework: Scan, Shield, Steer", "publisher": "Infosys", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists Scan/Shield/Steer components including Watchtower, maturity and risk assessments, Responsible AI Audit and Control Center." }, { "id": "s2", "url": "https://www.infosys.com/newsroom/press-releases/2024/enterprises-navigate-regulatory-ethical-complexities.html", "title": "Infosys Topaz Unveils Responsible AI Suite of Offerings", "publisher": "Infosys", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Announces the suite as 10+ offerings within Infosys Topaz and describes the Scan pillar's compliance single source of truth." } ], "claims": [ { "id": "c1", "statement": "Infosys describes the Topaz Responsible AI Suite as 10+ offerings built around a Scan, Shield and Steer framework, including a Responsible AI Watchtower for regulatory and threat monitoring, maturity and risk assessments, a Responsible AI Audit, and a Responsible AI Control Center.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Offering names and one-line scopes only; no documentation of tooling, control coverage or audit methodology.", "capability": "suite-composition" }, { "id": "c2", "statement": "Infosys' 2024 press release announces the Responsible AI Suite as part of Infosys Topaz, describing solutions that identify AI risk posture and legal obligations and generate a single source of truth for the compliance status of AI projects.", "support": "independently_supported", "sourceIds": [ "s2" ], "limitations": "Company newsroom release; corroborates existence and launch positioning, not delivered outcomes.", "capability": "launch-and-positioning" } ], "limitations": [ "Boundary between consulting services and Infosys-built software assets is unclear", "No evidence of third-party assurance or certification" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "IN", "US", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Hybrid service/product characteristics; classified as service because the suite is sold as delivery offerings. TCS' comparable Responsible AI portfolio was held for thinner offering-level evidence." }, { "name": "Inspect AI", "id": "inspect-ai", "companyName": "UK AI Security Institute", "companyUrl": "https://www.aisi.gov.uk", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI safety researcher", "Model evaluation lead" ], "problemsSolved": [ "standardised model and agent evaluations", "sandboxed execution of untrusted model code", "tool-use and multi-turn benchmark testing" ], "sources": [ { "id": "s1", "url": "https://inspect.aisi.org.uk/", "title": "Inspect AI", "publisher": "UK AI Security Institute", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Inspect is developed by the UK AI Security Institute and Meridian Labs; documents evaluation scope, datasets, flexible tool calling including MCP, multi-agent primitives, external agents, sandboxing options, and log reading." }, { "id": "s2", "url": "https://github.com/UKGovernmentBEIS/inspect_ai", "title": "Inspect: A framework for large language model evaluations", "publisher": "GitHub / UK AI Security Institute", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Repository under UKGovernmentBEIS, MIT license, created by the UK AI Security Institute, with facilities for prompt engineering, tool usage, multi-turn dialog and model-graded evaluations and over 200 pre-built evaluations." } ], "claims": [ { "id": "c1", "statement": "Inspect supports evaluations measuring coding, agentic tasks, reasoning, knowledge, behaviour and multi-modal understanding, with datasets of labelled samples loaded from Hugging Face, CSV, JSON or memory.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Does not document human feedback collection, red-team probe libraries or production monitoring.", "capability": "evaluation-scope" }, { "id": "c2", "statement": "Documented features include flexible tool calling with custom and MCP tools, built-in bash, python, web browsing and computer tools, multi-agent primitives, external agents such as Claude Code and Codex CLI, and sandboxed execution of untrusted model code in Docker, Kubernetes, Modal, Proxmox or Vagrant.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Tool-call approval is referenced in navigation but no policy-gating feature is described on the page fetched.", "capability": "tools-and-sandboxing" }, { "id": "c3", "statement": "The repository states Inspect was created by the UK AI Security Institute, is MIT licensed, and ships over 200 pre-built evaluations runnable against any model.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No enterprise tier, hosted deployment or support commitment is documented.", "capability": "ownership-and-license" } ], "limitations": [ "Research and assurance tooling: no RBAC, audit logging, retention control or immutable evidence store.", "Evaluation logs are local artefacts; nothing documented about tamper-evidence or long-term evidence retention." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "UK", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Model Context Protocol (MCP) tools", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Documented as a supported tool type; no vendor-specific MCP server integrations established." } ], "conflicts": [], "editorialNotes": "Repository remains under the UKGovernmentBEIS GitHub organisation while the framework is attributed to the UK AI Security Institute (with Meridian Labs named on the docs site). DEDUP FLAG: also present in services-oss.json. Recommend keeping one entry under evaluation (model/agent evaluation framework). Also surfaced in the open-source pass; evaluation kept as primary." }, { "name": "JDLA C認証 (AI Governance Core Certification)", "id": "jdla-c-certification", "companyName": "Japan Deep Learning Association (JDLA)", "companyUrl": "https://www.jdla.org", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Head of Compliance", "Corporate planning lead", "Chief AI Officer" ], "problemsSolved": [ "demonstrable AI governance status in Japan", "AI governance structure build-out", "third-party review" ], "sources": [ { "id": "s1", "url": "https://www.jdla.org/ai-governance/certification/", "title": "C認証｜AIガバナンス第三者認証制度", "publisher": "JDLA", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes entity-level third-party certification, two-year validity, review by JDLA's certification committee and the role of accredited consulting firms." }, { "id": "s2", "url": "https://www.jdla.org/news/20260804001/", "title": "AIガバナンス第三者認証制度「C認証」運用開始", "publisher": "JDLA", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Announces the scheme's launch with JDLA as certification body and describes it as reviewing AI management and operational structures." } ], "claims": [ { "id": "c1", "statement": "JDLA states C認証 certifies an organisation's AI governance structure at legal-entity level rather than individual AI tools or systems, is open to all corporate bodies including public entities, and is valid for two years with renewal every two years.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Detailed review criteria, fees and reviewer qualifications are not published on the page; content is Japanese-language only.", "capability": "scheme-scope" }, { "id": "c2", "statement": "JDLA states applications are reviewed by its certification committee, that JDLA-accredited consulting firms support organisations in building governance readiness, and that those consulting firms do not also perform the review.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Independence rule is stated by JDLA; no external oversight or accreditation of the scheme itself is evidenced.", "capability": "review-and-separation-of-duties" } ], "limitations": [ "Japanese market only; no English-language scheme documentation verified", "Scheme is young (launched 2026) with limited public certificate registry evidence" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "JP" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Launch announcement dated 2026-08-04 per JDLA news page; publishedAt left null because the page date was not independently confirmed in machine-readable form." }, { "name": "Knostic", "id": "knostic", "companyName": "Knostic", "companyUrl": "https://www.knostic.ai", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "identity-access", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "IAM manager", "Data governance lead" ], "problemsSolved": [ "LLM oversharing of sensitive content", "need-to-know access for AI assistants", "AI readiness assessment" ], "sources": [ { "id": "s1", "url": "https://www.knostic.ai/the-genai-knowledge-security-platform/knowledge-controls", "title": "Real-Time Knowledge Controls", "publisher": "Knostic", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-06-02", "supports": "Describes knowledge mapping, need-to-know evaluation via a policy-aware knowledge graph checking permissions and Purview labels, flagging of overshared content, and answer reshaping." }, { "id": "s2", "url": "https://www.knostic.ai/what-we-do", "title": "What We Do :: Knostic AI :: Enterprise AI Security", "publisher": "Knostic", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-08-24", "supports": "Describes need-to-know based access controls, LLM data-leak prevention, prompt-based discovery of overshared sensitive business content, and a Copilot Readiness Assessment." } ], "claims": [ { "id": "c1", "statement": "A policy-aware knowledge graph evaluates whether information should be surfaced by checking permissions, Microsoft Purview labels and custom rules.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page, not documentation; does not describe how label and permission data are ingested or refreshed.", "capability": "need-to-know-evaluation" }, { "id": "c2", "statement": "The platform identifies overexposed information reachable through enterprise LLMs and flags overshared content so access controls can be adjusted before AI surfaces it.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "No independent measurement of detection coverage; sources do not state which LLM assistants beyond Microsoft 365 Copilot are supported for enforcement.", "capability": "oversharing-detection" }, { "id": "c3", "statement": "Offers a Copilot Readiness Assessment that uses prompts built on a corpus of sensitive business topics to discover overshared content for specific user profiles.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Efficiency figures cited on the page are vendor-reported and unverified.", "capability": "copilot-readiness-assessment" } ], "limitations": [ "No public technical documentation was reachable in this session, so all capability claims rest on vendor product pages.", "Enforcement scope outside Microsoft 365 content is unclear." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "IL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Microsoft Purview sensitivity labels", "type": "announced", "sourceIds": [ "s1" ], "limitations": "Page states the knowledge graph checks Purview labels but gives no configuration or API documentation." } ], "conflicts": [], "editorialNotes": "Fits data-governance rather than identity-access because the control object is content exposure through LLM answers, not agent credentials." }, { "name": "Kong AI Gateway (with MCP Traffic Gateway)", "id": "kong-ai-gateway", "companyName": "Kong Inc.", "companyUrl": "https://konghq.com/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "API platform owner", "Platform engineering lead", "Enterprise architect" ], "problemsSolved": [ "governing remote MCP traffic", "turning APIs into governed agent tools", "LLM traffic routing and limits" ], "sources": [ { "id": "s1", "url": "https://developer.konghq.com/mcp/", "title": "MCP Traffic Gateway | Kong Docs", "publisher": "Kong Inc.", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes a trust and control layer for MCP traffic, API-to-MCP conversion via the AI MCP Proxy plugin, AI MCP OAuth2 authentication, access controls and rate limiting." }, { "id": "s2", "url": "https://developer.konghq.com/ai-gateway/", "title": "AI Gateway - Kong Docs", "publisher": "Kong Inc.", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes AI Gateway as a connectivity and governance layer for LLM, MCP and A2A traffic with provider routing, failover, streaming and prompt-inspecting AI policies." } ], "claims": [ { "id": "c1", "statement": "Kong AI Gateway acts as a trust and control layer for proxying traffic to MCP servers, enforcing access controls for MCP tool usage and securing MCP endpoints with the AI MCP OAuth2 plugin.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The page does not document authorization policy syntax, per-tool consent, or human approval steps.", "capability": "mcp-traffic-governance" }, { "id": "c2", "statement": "The AI MCP Proxy plugin converts API schemas into MCP tool definitions and aggregates multiple APIs into a single MCP server endpoint without requiring an LLM.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Conversion behaviour for complex or streaming APIs is not detailed.", "capability": "api-to-mcp-conversion" }, { "id": "c3", "statement": "AI Gateway policies can inspect request bodies so the gateway can act on the prompts passing through it, alongside routing, failover and load balancing across LLM providers.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Prompt inspection is described generically; no prompt-injection detector or safety classifier is documented on this page.", "capability": "prompt-aware-policies" } ], "limitations": [ "No documented LLM safety guardrails or DLP on the fetched pages", "Some capabilities are tied to the Konnect control plane" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas", "hybrid" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "MCP Traffic Gateway is documented as part of the AI Gateway product family, so it is recorded as one entry rather than two." }, { "name": "KPMG AI Trust services", "id": "kpmg-ai-trust-services", "companyName": "KPMG", "companyUrl": "https://kpmg.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "agent-discovery" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Chief Risk Officer", "Head of Internal Audit", "CISO" ], "problemsSolved": [ "AI inventory and risk assessment", "AI model validation", "independent AI attestation" ], "sources": [ { "id": "s1", "url": "https://kpmg.com/xx/en/what-we-do/services/kpmg-trusted-ai/ai-trust-services.html", "title": "AI Trust services", "publisher": "KPMG International", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Tabulates AI readiness/compliance, AI risk transformation and AI risk monitoring components including AI assurance and AI system cards." }, { "id": "s2", "url": "https://kpmg.com/us/en/media/news/kpmg-expands-ai-trust-services-with-new-ai-assurance-capabilities.html", "title": "KPMG expands AI Trust services with new AI Assurance capabilities", "publisher": "KPMG US", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Announces AI Assurance components: model risk assessments, control testing, model validation, real-time systems assessments, and assurance/attestation against SOC, FedRAMP, SWIFT and HITRUST." } ], "claims": [ { "id": "c1", "statement": "KPMG lists AI Trust service components including AI risk assessment and strategy, AI regulation and compliance, AI systems inventory, AI governance and policy implementation, AI security and privacy, AI assurance, and AI system cards.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Service catalogue page; no methodology, staffing model, or deliverable samples published.", "capability": "service-components" }, { "id": "c2", "statement": "KPMG's release states new AI Assurance capabilities include AI model validation for accuracy, assumptions and regulatory compliance, real-time systems assessments, and formal independent assurance over AI systems against defined standards including SOC, FedRAMP, SWIFT and HITRUST.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "A firm press release, not an accreditation record; it does not evidence which attestation standards KPMG is licensed to issue for AI-specific scopes.", "capability": "assurance-and-attestation" } ], "limitations": [ "Attestation scope wording comes from a firm release, not a regulator or accreditation body", "No pricing or engagement scope published" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "KPMG references enabling technology from Microsoft, ServiceNow and Cranium; not recorded as integrations because the release only names platforms, not documented product integrations." }, { "name": "Laminar", "id": "laminar", "companyName": "Laminar", "companyUrl": "https://lmnr.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineer", "Agent platform developer" ], "problemsSolved": [ "tracing long agent runs and sub-agents", "run recording and replay for debugging", "dataset creation from traces", "CI evaluation of agents" ], "sources": [ { "id": "s1", "url": "https://docs.lmnr.ai/", "title": "Laminar overview", "publisher": "Laminar", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Laminar as an open-source OpenTelemetry-native observability and debugging platform for AI agents tracing every LLM call, tool call and sub-agent, transcript-style trace view, datasets from production traces, evals locally or in CI, playground span replay and a recording debugger." }, { "id": "s2", "url": "https://github.com/lmnr-ai/lmnr", "title": "lmnr-ai/lmnr: Laminar", "publisher": "GitHub / Laminar", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Apache-2.0 license, lmnr-ai owner, OpenTelemetry-based automatic tracing powered by OpenLLMetry, captured trace input/output, latency, cost and token count, observe decorator, image tracing, managed platform at lmnr.ai and documented self-hosting." } ], "claims": [ { "id": "c1", "statement": "Laminar traces every LLM call, tool call and sub-agent an agent runs, supports agents working through hundreds of steps across parallel sub-agents, and presents each trace as a transcript of agent inputs, LLM turns, tool calls and sub-agents rather than a span tree.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Sessions, human feedback and cost/token telemetry are not stated on the overview page, though the repository lists cost and token count as captured trace data.", "capability": "agent-trace-capture" }, { "id": "c2", "statement": "The debugger records a run and replays it, and the playground can replay any traced span with prompts or models swapped for side-by-side comparison; evals can be run against datasets locally or in CI to catch regressions before shipping.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Replay is a developer debugging affordance; nothing documents tamper-evident storage or retention of replayed runs.", "capability": "record-and-replay" }, { "id": "c3", "statement": "The repository states Laminar is Apache-2.0 licensed under the lmnr-ai organization, offers OpenTelemetry-based automatic tracing powered by OpenLLMetry with two lines of code, captures trace input/output, latency, cost and token count, and documents self-hosting alongside the managed platform.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No audit logs, RBAC, SSO, retention or compliance features are stated; latest release listed predates 2026.", "capability": "license-and-instrumentation" } ], "limitations": [ "No documented multi-user access control, audit trail, retention policy or data masking.", "Developer debugging orientation: replay serves engineering diagnosis, not regulated incident evidence." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "OpenTelemetry", "type": "native", "sourceIds": [ "s1", "s2" ], "limitations": "Described as OpenTelemetry-native via OpenLLMetry; semantic-convention conformance not documented." } ], "conflicts": [], "editorialNotes": "Hybrid: Apache-2.0 self-hostable stack plus managed cloud at lmnr.ai." }, { "name": "Langfuse", "id": "langfuse", "companyName": "Langfuse", "companyUrl": "https://langfuse.com", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI platform lead", "AI engineer", "Compliance-minded engineering manager" ], "problemsSolved": [ "tracing LLM and agent executions", "cost and latency analysis", "human annotation and LLM-as-judge scoring", "dataset-based experiments" ], "sources": [ { "id": "s1", "url": "https://langfuse.com/docs", "title": "Langfuse Overview", "publisher": "Langfuse", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Langfuse is an open-source AI engineering platform; traces include all LLM and non-LLM calls, retrieval, embedding and API calls; sessions track multi-turn conversations and agentic workflows; cost and latency analysis; prompt versioning; datasets and experiments; user feedback, annotation queues, LLM-as-a-judge, code evaluators and production-trace scoring." }, { "id": "s2", "url": "https://langfuse.com/docs/administration/rbac", "title": "Access Control (RBAC)", "publisher": "Langfuse", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents users, organizations, projects and roles (Owner, Admin, Member, Viewer, None), inheritance and override of organization roles by project roles, and restrictions on assigning roles above one's own." }, { "id": "s3", "url": "https://langfuse.com/pricing", "title": "Pricing", "publisher": "Langfuse", "type": "official_pricing", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists audit logs and SCIM API for Enterprise, fine-grained RBAC in the Teams add-on, enterprise SSO and SSO enforcement, data retention management, client-side data masking, SOC2 Type II and ISO27001 reports, HIPAA-ready region, GDPR DPA, AWS PrivateLink, and US/EU/JP data regions." } ], "claims": [ { "id": "c1", "statement": "Traces include all LLM and non-LLM calls such as retrieval, embedding and API calls, support tracking multi-turn conversations as sessions plus user tracking, and can represent agents as graphs, with cost and latency shown in a timeline view.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs overview does not explicitly name tool calls or multi-agent handoffs as distinct span types.", "capability": "trace-and-session-capture" }, { "id": "c2", "statement": "Access control is organised around users, organizations and projects with Owner, Admin, Member, Viewer and None roles, project-level roles overriding organization roles, and users only able to assign roles at or below their own.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The RBAC page does not state which plans include fine-grained project roles.", "capability": "rbac" }, { "id": "c3", "statement": "Pricing documents audit logs and SCIM for Enterprise, organization- and project-level plus fine-grained RBAC, enterprise SSO with enforcement, data retention management, client-side data masking, SOC2 Type II and ISO27001 reports, a GDPR data processing agreement, and US, EU, JP or HIPAA data regions.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Pricing-page feature lists do not describe audit-log contents, immutability or retention granularity, and compliance reports are referenced rather than published.", "capability": "governance-controls" }, { "id": "c4", "statement": "Langfuse runs managed LLM-as-a-judge evaluations on production or development traces, supports code evaluators, custom pipelines, manual labelling and human annotation queues, and dataset experiments.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Judge accuracy and calibration are not addressed.", "capability": "evaluation-features" } ], "limitations": [ "Audit logging is an Enterprise-tier feature and its scope, immutability and export are not documented.", "Trace records are operational telemetry; nothing documents write-once storage suitable as regulatory evidence." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "EU", "US", "JP", "global" ], "pricing": { "model": "tiered subscription (Hobby, Core, Pro, Enterprise) with a Teams add-on", "url": "https://langfuse.com/pricing" }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Strongest governance evidence in this batch (audit logs, RBAC, retention, masking, data regions). Hybrid: open-source core plus managed cloud tiers." }, { "name": "LangSmith", "id": "langsmith", "companyName": "LangChain", "companyUrl": "https://www.langchain.com", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI engineering lead", "Platform engineer" ], "problemsSolved": [ "debugging agent failures from traces", "production quality monitoring", "dataset building from traces", "annotation and feedback collection" ], "sources": [ { "id": "s1", "url": "https://docs.smith.langchain.com/", "title": "LangSmith Observability", "publisher": "LangChain", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States LangSmith Observability provides full visibility into LLM applications, traces are the record of what agents did in production, used to debug failures, monitor quality and build evaluation datasets, with annotation queues, inline annotation and online evaluation rules." }, { "id": "s2", "url": "https://docs.langchain.com/langsmith/administration-overview", "title": "Administration overview", "publisher": "LangChain", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents managing users, organizations and workspaces; RBAC restricted to Enterprise; Workspace Admin/Editor/Viewer system roles; organization roles and custom role creation by organization admins." } ], "claims": [ { "id": "c1", "statement": "LangSmith Observability provides visibility from individual traces to production-wide performance metrics, with traces described as the record of what agents did in production, plus annotation of outputs, feedback via queues or inline annotation, and rules for online evaluations.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The overview does not state whether tool calls, sessions, token counts or costs are captured as structured fields.", "capability": "trace-capture-and-feedback" }, { "id": "c2", "statement": "RBAC is available only to Enterprise customers, with other plans defaulting to the Admin role for all users; built-in workspace roles are Workspace Admin, Editor and Viewer, and organization admins can create custom roles.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Audit logs, retention configuration, PII redaction and compliance certifications are not documented on the pages fetched.", "capability": "rbac" } ], "limitations": [ "No documented audit trail, retention policy or data-masking control, so trace records are developer telemetry rather than governance evidence.", "Self-hosted deployment was not established from the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Docs now served from docs.langchain.com; the smith.langchain.com docs path redirects into the same documentation set." }, { "name": "LangSmith Deployment (LangGraph)", "id": "langsmith-deployment-langgraph", "companyName": "LangChain", "companyUrl": "https://www.langchain.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineering lead", "Platform engineer", "Head of AI" ], "problemsSolved": [ "running long-lived agents reliably", "human approval of agent actions", "agent deployment lifecycle" ], "sources": [ { "id": "s1", "url": "https://docs.langchain.com/langgraph-platform/index", "title": "LangSmith Deployment - Docs by LangChain", "publisher": "LangChain", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes LangSmith Deployment as a workflow orchestration runtime for agent workloads with durable execution, streaming, horizontal scaling, Cloud/standalone/self-hosted options, assistants-threads-runs model, framework-agnostic deployment and a private preview API for Managed Deep Agents." }, { "id": "s2", "url": "https://docs.langchain.com/oss/python/langchain/human-in-the-loop", "title": "Responding to interrupts", "publisher": "LangChain", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents human-in-the-loop middleware that adds oversight to tool calls, interrupts execution per configurable policy, persists graph state via LangGraph persistence, and supports approve, edit, reject and respond decisions." } ], "claims": [ { "id": "c1", "statement": "Human-in-the-loop middleware checks each tool call against a configurable policy and interrupts execution so a human can approve, edit, reject with feedback, or respond directly, with graph state persisted so the run resumes later.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Approval policy is authored in code by developers; the docs do not describe an administrator-facing approval console or segregation of duties.", "capability": "human-in-the-loop-approvals" }, { "id": "c2", "statement": "The Agent Server runtime provides durable execution, streaming and horizontal scaling with assistants for configuration, threads for state and runs for workloads, and can run in LangChain Cloud, a standalone server or self-hosted.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Page does not document tenant isolation guarantees or audit-log retention.", "capability": "durable-agent-runtime" }, { "id": "c3", "statement": "Deployment is framework-agnostic, supporting LangGraph, LangChain, Deep Agents, Google ADK, Claude Agent SDK, Strands, CrewAI and AutoGen agents.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Managed Deep Agents are described as available through a private preview API, so not all listed paths are generally available.", "capability": "framework-agnostic-deployment" }, { "id": "c4", "statement": "Administrative governance features such as role-based permission scoping, audit logging and tenant isolation are not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1", "s2" ], "limitations": "Absence of evidence, not evidence of absence; LangSmith may document these elsewhere.", "capability": "versioning-and-permission-scoping" } ], "limitations": [ "Governance surface is developer-only: approval policies are code, not tenant-level administrative policy, in the pages reviewed.", "No audit-log, RBAC or tenant-isolation documentation was found in this session.", "Naming has shifted from LangGraph Platform to LangSmith Deployment, which complicates procurement references." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "LangGraph Platform" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Deduped LangGraph (OSS library), LangGraph Platform and LangSmith Deployment into one entry because they are one product line with a shared runtime; recorded LangGraph Platform as a former name." }, { "name": "LangWatch", "id": "langwatch", "companyName": "LangWatch", "companyUrl": "https://langwatch.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI platform lead", "AI engineer", "Compliance-minded engineering manager" ], "problemsSolved": [ "tracing LLM calls and tool usage", "multi-turn agent testing with simulated users", "prompt versioning and optimisation", "data-sovereign deployment of AI telemetry" ], "sources": [ { "id": "s1", "url": "https://docs.langwatch.ai/", "title": "LangWatch: The Complete LLMOps Platform", "publisher": "LangWatch", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States every LLM call, tool usage and user interaction is tracked with detailed traces, spans and metadata; describes evaluations, datasets and experiments, agent testing with a simulated user in multi-turn conversations, prompt versioning, per-user workspaces and an AI Gateway with budgets, routing policies and anomaly detection." }, { "id": "s2", "url": "https://docs.langwatch.ai/self-hosting/overview", "title": "Self-hosting overview - LangWatch", "publisher": "LangWatch", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents self-hosting for data sovereignty, regulatory compliance (GDPR, SOC 2) or air-gapped environments with a single build; audit logs showing who did what and when; role-based access control with fine-grained permissions across projects and teams; SSO and SCIM; Cloud Enterprise dedicated instances; hybrid control-plane/data-plane split." } ], "claims": [ { "id": "c1", "statement": "LangWatch automatically tracks every LLM call, tool usage and user interaction with detailed traces, spans and metadata, showing the full conversation flow, and supports agent testing of multi-turn conversations with a simulated user.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Multi-agent handoffs and human feedback capture are not documented on the introduction page.", "capability": "trace-capture" }, { "id": "c2", "statement": "Self-hosting documentation states the same software as LangWatch Cloud can be self-hosted for data sovereignty, regulatory compliance or air-gapped environments with no separate community or enterprise build, and lists audit logs giving visibility into who did what and when, role-based access control with fine-grained permissions across projects and teams, and SSO with SCIM provisioning.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Retention configuration and PII redaction are not documented on that page, and audit-log contents and immutability are not specified.", "capability": "self-hosting-and-governance" }, { "id": "c3", "statement": "Documented deployment models include self-hosted, Cloud Enterprise with dedicated instances in a chosen cloud region, and a hybrid model where LangWatch runs the control plane while ClickHouse and S3 data planes sit in the customer VPC.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Compliance references (GDPR, SOC 2) appear as motivations for self-hosting rather than as certifications of the vendor.", "capability": "deployment-models" }, { "id": "c4", "statement": "The introduction describes an AI Gateway with budgets, routing policies and anomaly detection, and evaluations positioned to guard against harm.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented enforcement semantics, policy language or blocked-call logging.", "capability": "gateway-policies" } ], "limitations": [ "Audit logs and RBAC are documented as features of the self-hosted/enterprise editions without detail on log contents, immutability or retention.", "Simulation and evaluation features overlap with dedicated eval vendors; depth per feature is thinner than in single-purpose tools." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "hybrid" ], "geographies": [ "EU", "NL", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Netherlands-based vendor; classified observability-primary with evaluation secondary because tracing is the lead documented capability and simulation/eval sit alongside it." }, { "name": "LlamaFirewall", "id": "llamafirewall", "companyName": "Meta", "companyUrl": "https://ai.meta.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "evaluation" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI engineer", "Application security engineer", "AI safety researcher" ], "problemsSolved": [ "jailbreak and injection scanning for agents", "detecting goal misalignment in agent reasoning", "screening agent-generated code" ], "sources": [ { "id": "s1", "url": "https://meta-llama.github.io/PurpleLlama/LlamaFirewall/", "title": "LlamaFirewall", "publisher": "Meta", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Shows LlamaFirewall configuration with PROMPT_GUARD scanners for user and system roles, runtime scan calls, and an email agent example detecting and blocking malicious prompt injections." }, { "id": "s2", "url": "https://ai.meta.com/research/publications/llamafirewall-an-open-source-guardrail-system-for-building-secure-ai-agents/", "title": "LlamaFirewall: An open source guardrail system for building secure AI agents", "publisher": "Meta", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes LlamaFirewall as an open-source guardrail framework used in production at Meta, with PromptGuard 2, Agent Alignment Checks (experimental) and CodeShield." } ], "claims": [ { "id": "c1", "statement": "LlamaFirewall provides PromptGuard 2 as a jailbreak detector, Agent Alignment Checks as a chain-of-thought auditor for prompt injection and goal misalignment, and CodeShield as an online static analysis engine for insecure generated code.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Published as research and open source; production support commitments are not stated.", "capability": "layered-agent-scanners" }, { "id": "c2", "statement": "The framework is designed for integration into existing AI agents and LLM applications, with scanners configured per role and invoked on input at runtime.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The documentation page fetched shows prompt-guard scanning only and does not demonstrate tool-call or output enforcement.", "capability": "runtime-scanning-integration" }, { "id": "c3", "statement": "Meta states Agent Alignment Checks show stronger efficacy against indirect injections than prior approaches but remain experimental.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No accuracy figures for enterprise workloads are provided.", "capability": "experimental-limitation" } ], "limitations": [ "No managed service or dashboard", "Alignment checking described as experimental" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Part of Meta's Purple Llama project; recorded as a distinct product from Prompt Guard models, which are components of it." }, { "name": "Lumenova AI Platform", "id": "lumenova-ai", "companyName": "Lumenova AI", "companyUrl": "https://www.lumenova.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "evaluation", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief risk officer", "AI governance lead", "Model validator" ], "problemsSolved": [ "manual AI risk reviews", "fragmented compliance processes", "limited model transparency for business teams" ], "sources": [ { "id": "s1", "url": "https://www.lumenova.ai/platform/", "title": "Platform — Lumenova AI", "publisher": "Lumenova AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes automated risk-review workflows, 200+ quantitative and qualitative model metrics, embedded governance guardrails, monitoring and security layer with observability and guardrails, evaluations and prompt versioning." }, { "id": "s2", "url": "https://www.gov.uk/ai-assurance-techniques/lumenova-ai-governance-risk-management-and-compliance-platform", "title": "Lumenova AI Governance, Risk Management, and Compliance Platform", "publisher": "UK Government (DSIT)", "type": "other", "accessedAt": "2026-09-06", "publishedAt": "2024-04-09", "supports": "Describes the platform as SaaS for enterprise AI governance addressing safety/security/robustness, transparency, fairness and accountability across many sectors and AI use cases." } ], "claims": [ { "id": "c1", "statement": "Platform page states manual risk reviews and fragmented compliance processes are replaced by automated workflows and that AI models can be assessed across 200+ quantitative and qualitative metrics, with layers for assessment and governance, quality and operations, and monitoring and security including guardrails.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page; the metric list, computation method and which layers require separate modules are not evidenced.", "capability": "workflow-and-metric-assessment" }, { "id": "c2", "statement": "A UK Government AI assurance techniques entry describes Lumenova AI's Governance, Risk Management and Compliance Platform as a SaaS solution for governing AI systems at enterprise level, addressing safety, security and robustness, transparency and explainability, fairness and accountability across listed sectors and AI use cases.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Vendor-submitted case study published April 2024; it is not an independent evaluation and may predate current features.", "capability": "saas-enterprise-governance" } ], "limitations": [ "No product documentation was located; capability depth unverified.", "Vendor headquarters not stated on fetched pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept in ai-risk rather than evaluation because the evidenced framing is governance/risk workflow with evaluation as a component." }, { "name": "Maxim AI", "id": "maxim-ai", "companyName": "Maxim AI", "companyUrl": "https://www.getmaxim.ai", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI engineering lead", "Head of product quality" ], "problemsSolved": [ "multi-turn agent simulation", "evaluator-based scoring across versions", "human review workflows", "dataset curation from production interactions" ], "sources": [ { "id": "s1", "url": "https://www.getmaxim.ai/products/agent-simulation-evaluation", "title": "Agent Simulation Evaluation - Maxim AI", "publisher": "Maxim AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes simulating multi-turn interactions across scenarios and personas, scaling to thousands of scenarios, prebuilt and custom evaluators, evaluation dashboards across versions, human evaluation pipelines, CI/CD integration and dataset curation from real interactions and human feedback." }, { "id": "s2", "url": "https://www.getmaxim.ai/docs", "title": "Platform Overview - Maxim Docs", "publisher": "Maxim AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes an end-to-end platform for simulation, evaluation and observability of AI agents, real-time production logs in repositories, distributed tracing, prompt experimentation, evaluator types and human evaluation, and periodic automated quality checks on production logs." } ], "claims": [ { "id": "c1", "statement": "Maxim documents simulating multi-turn interactions across a wide range of scenarios and user personas, scaling testing across thousands of scenarios and test cases, and creating custom simulation environments.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Does not document how faithfully simulated personas represent real user populations.", "capability": "simulation-and-personas" }, { "id": "c2", "statement": "Evaluations can use AI, programmatic or statistical evaluators from an evaluator store or custom evaluators, visualised across versions and test suites, with human evaluation pipelines for last-mile checks and periodic quality checks on production logs.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "No documented immutable record of who approved a release based on those scores.", "capability": "evaluators-and-human-review" }, { "id": "c3", "statement": "The observability suite monitors real-time production logs organised into repositories per application, logged and analysed using distributed tracing.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Span taxonomy (LLM, tool, agent handoff) is not enumerated on the pages fetched.", "capability": "production-tracing" } ], "limitations": [ "No audit logging, RBAC/SSO, retention configuration or compliance certification documented on the pages fetched.", "Self-hosted deployment not documented, which may constrain regulated buyers." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "library" ], "geographies": [ "IN", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Positioned evaluation-primary (simulation and evaluation are the lead documented capabilities) with observability secondary." }, { "name": "Microsoft Agent 365", "id": "microsoft-agent-365", "companyName": "Microsoft", "companyUrl": "https://www.microsoft.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "runtime-controls", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Microsoft 365 administrator", "Security administrator", "AI administrator" ], "problemsSolved": [ "tenant agent registry", "shadow agent detection on endpoints", "agent governance actions" ], "sources": [ { "id": "s1", "url": "https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide", "title": "Manage agent registry in Microsoft 365 admin center", "publisher": "Microsoft", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the registry view of all agents available to the organization and the four agent types it distinguishes." }, { "id": "s2", "url": "https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-shadow-ai?view=o365-worldwide", "title": "Shadow AI in Microsoft 365 admin center (Preview)", "publisher": "Microsoft", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents preview detection and governance of unmanaged AI agents with prerequisites: Defender for Endpoint, E5, Intune, Global Secure Access and specific admin roles." }, { "id": "s3", "url": "https://www.microsoft.com/en-us/security/blog/2026/05/01/microsoft-agent-365-now-generally-available-expands-capabilities-and-integrations/", "title": "Microsoft Agent 365, now generally available, expands capabilities and integrations", "publisher": "Microsoft", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-05-01", "supports": "Announces Agent 365 general availability as a control plane to observe, govern and secure agents across apps, endpoints and cloud, including partner-built agents." } ], "claims": [ { "id": "c1", "statement": "The agent registry in the Microsoft 365 admin center gives a centralized view of all agents available to an organization, split into Microsoft agents, external partner-built agents, agents published by the org, and agents shared by their creator.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Registry lists agents available in the Microsoft ecosystem; agents on other platforms are not covered by this view.", "capability": "tenant-agent-registry" }, { "id": "c2", "statement": "A preview Shadow AI page detects and governs unmanaged AI agents such as OpenClaw, requiring Microsoft Defender for Endpoint for shadow agent detection, Microsoft 365 E5, Intune enrolment, and Global Secure Access for additional usage metadata.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Explicitly public preview under the Frontier programme; supported agents and behaviours may change, and detection requires the listed Microsoft stack.", "capability": "local-shadow-agent-detection" }, { "id": "c3", "statement": "Microsoft announced general availability of Agent 365 for commercial customers on 1 May 2026 as a control plane to observe, govern and secure agents, including agents from ecosystem partners.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Several capabilities announced alongside GA are described as previews; partner coverage depth is not specified.", "capability": "ga-control-plane" } ], "limitations": [ "Shadow agent detection is preview and dependent on the Microsoft endpoint/network stack", "Non-Microsoft agent platforms are covered only through partner integrations not verified here" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Dedup decision: the Microsoft 365 admin center Shadow AI preview page is folded into this entry as a claim rather than listed as a separate product, since it is a page within the same admin surface." }, { "name": "Microsoft Entra Agent ID", "id": "microsoft-entra-agent-id", "companyName": "Microsoft", "companyUrl": "https://www.microsoft.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "IAM manager", "Identity architect" ], "problemsSolved": [ "distinguishing agent actions from human and workload identities", "issuing and retiring agent identities at scale", "applying conditional access policy to agents" ], "sources": [ { "id": "s1", "url": "https://learn.microsoft.com/en-us/entra/agent-id/what-are-agent-identities", "title": "What are agent identities? - Microsoft Entra Agent ID", "publisher": "Microsoft", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-06-15", "supports": "Defines Entra Agent ID as the platform for creating and managing agent identities and blueprints, including bulk creation and retirement." }, { "id": "s2", "url": "https://learn.microsoft.com/en-us/entra/agent-id/whats-new-agent-id", "title": "What's new in Microsoft Entra Agent ID", "publisher": "Microsoft", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-05-01", "supports": "Lists agent-specific OAuth flows, admin center blueprint wizard, Auth SDK sidecar pattern and cascade cleanup / soft delete." }, { "id": "s3", "url": "https://www.microsoft.com/en-us/security/business/identity-access/microsoft-entra-agent-id", "title": "Secure Agent Access with Microsoft Entra", "publisher": "Microsoft", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States agents are treated as first-class identities governed by Conditional Access, identity governance and identity protection." } ], "claims": [ { "id": "c1", "statement": "Entra Agent ID provides a platform for creating and managing agent identities and agent identity blueprints as identity accounts in Entra ID.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The page does not quantify supported agent frameworks or state SLAs.", "capability": "agent-identity-issuance" }, { "id": "c2", "statement": "Organizations can create agent identities in bulk, apply consistent policies, and use automated cascade cleanup and soft-delete for agent identities.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Release notes list features without stating GA dates for each item.", "capability": "bulk-lifecycle-and-cleanup" }, { "id": "c3", "statement": "Agent identities are governed with existing Entra capabilities including Conditional Access, identity governance, identity protection and network controls.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "Marketing page; does not detail which Conditional Access conditions apply to agent tokens.", "capability": "conditional-access-for-agents" } ], "limitations": [ "No fetched source states pricing or licensing for agent identities.", "Discovery of agents built outside Microsoft platforms is not evidenced." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Distinct from Microsoft Agent 365 (agent management/ops), which the docs describe as a separate product. Also surfaced in agent-discovery research; identity-access kept as primary because the evidenced artifact is an agent identity object, with discovery as a secondary." }, { "name": "Microsoft Foundry Agent Service", "id": "microsoft-foundry-agent-service", "companyName": "Microsoft", "companyUrl": "https://www.microsoft.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "identity-access", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI platform lead", "Cloud architect", "Engineering manager" ], "problemsSolved": [ "hosting and scaling agents", "agent identity assignment", "multi-framework agent deployment" ], "sources": [ { "id": "s1", "url": "https://learn.microsoft.com/en-us/azure/foundry/agents/overview", "title": "What is Microsoft Foundry Agent Service?", "publisher": "Microsoft Learn", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-08-19", "supports": "Documents Foundry Agent Service as a managed platform for building, deploying and scaling agents, with prompt, hosted, Responses API and ephemeral paths, supported frameworks, and hosted-agent managed endpoint, scaling, dedicated Entra identity, session state and observability." }, { "id": "s2", "url": "https://learn.microsoft.com/en-us/agent-framework/overview/", "title": "Microsoft Agent Framework Overview", "publisher": "Microsoft Learn", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-08-25", "supports": "Documents Microsoft Agent Framework agents and workflows, provider support including Microsoft Foundry, and Harness Agent capabilities including planning, context compaction, memory, don't-ask-again tool approval and observability." } ], "claims": [ { "id": "c1", "statement": "Offers prompt agents run with no code or infrastructure, hosted agents supplied as a container image or zipped source, direct Responses API calls, and ephemeral agents defined in application code.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation does not state per-path regional availability or quota limits.", "capability": "managed-agent-hosting" }, { "id": "c2", "statement": "Hosted agents are provided with a managed endpoint, automatic scaling, a dedicated Microsoft Entra identity, session-level state persistence and end-to-end observability.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The overview does not detail how Entra permissions are scoped per agent or how least privilege is enforced.", "capability": "per-agent-entra-identity" }, { "id": "c3", "statement": "Hosted agents can be built with Agent Framework, LangGraph, the OpenAI Agents SDK, the Anthropic Agent SDK, the GitHub Copilot SDK or custom code.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Microsoft Agent Framework is documented separately; the overview does not state version support matrices.", "capability": "framework-support" }, { "id": "c4", "statement": "Microsoft Agent Framework's Harness Agent documents 'don't-ask-again' tool approval alongside planning, context compaction, memory and observability.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "This is a framework feature exercised in code, not a tenant-level approval workflow in the Agent Service.", "capability": "tool-approval-in-framework" }, { "id": "c5", "statement": "Administrative audit logging and role-based permission scoping for agent actions are not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1" ], "limitations": "Absence of evidence only; Azure platform-level RBAC and logging likely exist but were not verified here.", "capability": "audit-logging-and-rbac" } ], "limitations": [ "Governance evidence is limited to identity assignment and observability; approvals are a framework-level developer feature.", "No documented audit trail of agent actions or admin approval gate was verified in this session.", "Product naming has changed repeatedly (Azure AI Agent Service to Foundry Agent Service), so older references may not match current docs." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "Azure AI Foundry Agent Service", "Azure AI Agent Service" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Microsoft Agent Framework was not given its own entry to avoid double-counting one Microsoft agent stack; it is cited as the framework layer beneath the hosted service." }, { "name": "Microsoft Purview Data Security Posture Management for AI", "id": "microsoft-purview-dspm-for-ai", "companyName": "Microsoft", "companyUrl": "https://www.microsoft.com", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Data protection officer", "Microsoft 365 administrator" ], "problemsSolved": [ "AI data oversharing", "sensitive data in AI prompts", "shadow AI usage visibility" ], "sources": [ { "id": "s1", "url": "https://learn.microsoft.com/en-us/purview/dspm-for-ai", "title": "Learn about Data Security Posture Management for AI", "publisher": "Microsoft Learn", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2025-12-15", "supports": "Documents DSPM for AI reports, Apps and agents dashboard, ready-to-use prompt DLP policies, data risk assessments, and browser-extension prerequisite for third-party AI site discovery." }, { "id": "s2", "url": "https://learn.microsoft.com/en-us/purview/ai-microsoft-purview", "title": "Microsoft Purview data security and compliance protections for Microsoft 365 Copilot and other generative AI apps", "publisher": "Microsoft Learn", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-05-27", "supports": "Lists the Copilot, enterprise AI and other AI app categories in scope and states agents inherit parent-app security and compliance capabilities." } ], "claims": [ { "id": "c1", "statement": "Provides reports on Copilot experiences and agents, enterprise AI apps and other AI apps, including an Apps and agents dashboard showing sensitive data accessed by each agent.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation does not quantify detection accuracy or state which agent frameworks outside Microsoft are inventoried.", "capability": "ai-activity-reporting" }, { "id": "c2", "statement": "Includes data risk assessments to identify, remediate and monitor potential oversharing of data to AI apps.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Source does not describe remediation guarantees or scope beyond Microsoft 365 content.", "capability": "oversharing-risk-assessment" }, { "id": "c3", "statement": "Discovery of user activity on third-party generative AI sites requires the Purview browser extension and devices onboarded to Purview.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Implies no coverage for unmanaged or non-onboarded devices.", "capability": "third-party-ai-discovery" }, { "id": "c4", "statement": "Where AI apps support agents, those agents inherit the same security and compliance capabilities as the parent AI app.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Source lists supported AI app categories but does not enumerate per-agent control differences.", "capability": "agent-inheritance-of-controls" } ], "limitations": [ "The primary documentation page is labelled '(classic)', so feature parity with any newer DSPM for AI experience is unclear.", "Depth of coverage for non-Microsoft agent runtimes is not established by the sources fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "AI Hub" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept as a single entry rather than splitting Purview DLP, Insider Risk and DSPM for AI; only the AI-specific module is in scope for this category." }, { "name": "Mindgard", "id": "mindgard", "companyName": "Mindgard", "companyUrl": "https://mindgard.ai", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "agent-security" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AI security engineer" ], "problemsSolved": [ "automated AI red teaming", "agent and guardrail security testing", "AI attack surface enumeration" ], "sources": [ { "id": "s1", "url": "https://mindgard.ai/", "title": "Mindgard - Automated AI Red Teaming & AI Security Platform", "publisher": "Mindgard", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes discovering exploits, assessing risk and defending AI systems and agents, attacker-style profiling of models, agents, tools and behaviours, attack surface enumeration, agent security testing, guardrail testing and AI risk compliance reporting." }, { "id": "s2", "url": "https://docs.mindgard.ai/", "title": "Mindgard Docs", "publisher": "Mindgard", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documentation index offering a quickstart, demo model testing, CLI testing, attack library, remediations, SDK and command-line reference, and an ever-growing list of LLM/ML attack techniques." } ], "claims": [ { "id": "c1", "statement": "Mindgard documents testing applications and models against an ever-growing list of LLM and ML attack techniques, with a quickstart, CLI-based testing, attack library, remediations and SDK reference.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The documentation index does not list specific attack categories, scoring methodology or evidence retention.", "capability": "attack-execution" }, { "id": "c2", "statement": "The product page states Mindgard profiles AI systems the way attackers do by mapping models, agents, tools and behaviours before attack execution, and lists AI attack surface enumeration, agent evaluation and security scanning, guardrail testing and AI risk compliance reporting.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page; no reference documentation was found for the compliance reporting or runtime protection claims.", "capability": "profiling-and-reporting" } ], "limitations": [ "No documented audit logs, RBAC/SSO, retention configuration or compliance certifications on the pages fetched.", "Compliance-reporting capability is claimed on marketing pages only, so its governance value is unverified." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "UK", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept in evaluation (adversarial testing) with agent-security secondary; overlaps with security-category research, flagged for dedup at merge. DEDUP FLAG: also present in runtime-security.json. Recommend keeping the evaluation entry (documented scope is adversarial testing/red teaming, not runtime enforcement) and dropping the duplicate. Also surfaced in agent-security research; evaluation kept as primary because the evidenced function is automated adversarial testing rather than runtime blocking." }, { "name": "MintMCP", "id": "mintmcp", "companyName": "MintMCP", "companyUrl": "https://www.mintmcp.com/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "identity-access", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "IT operations lead", "Platform engineer" ], "problemsSolved": [ "approving and restricting MCP servers", "per-agent credentials and revocation", "tool call audit and behavioural rules" ], "sources": [ { "id": "s1", "url": "https://www.mintmcp.com/", "title": "MintMCP - MCP Gateway & AI Agent Infrastructure", "publisher": "MintMCP", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the gateway decides which tools an agent reaches, provides role-based endpoints, agent identities with own credentials, SSO via SAML/OIDC and per-call traceability." }, { "id": "s2", "url": "https://docs.mintmcp.com/", "title": "Introduction to MintMCP", "publisher": "MintMCP", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents availability as a managed cloud service, access policies and role-based tool sets, per-agent credentials, tool call logging, and Agent Monitor hook rules to detect and block risky behaviour." } ], "claims": [ { "id": "c1", "statement": "MintMCP enforces access policies and role-based tool sets through a single gateway, with tool-level switches that disable destructive actions.", "support": "documented", "sourceIds": [ "s2", "s1" ], "limitations": "Documentation does not describe policy expression syntax or fine-grained argument-level conditions.", "capability": "role-scoped-tool-access" }, { "id": "c2", "statement": "Autonomous agents receive their own delegated credentials and audit trail rather than borrowing human logins, and can be revoked individually.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Revocation propagation time and credential storage details are not documented.", "capability": "per-agent-credentials" }, { "id": "c3", "statement": "An Agent Monitor uses hook scripts to capture file reads, command execution and MCP tool calls, and lets users create rules that detect and block risky behaviour.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The detection library is user-authored; no built-in prompt-injection classifier is documented.", "capability": "behaviour-blocking-rules" } ], "limitations": [ "Self-hosted deployment is not documented, only offered via sales contact", "Vendor company details beyond the product name are not disclosed on fetched pages" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Company and product share the name; no parent company evidence found." }, { "name": "MLflow Model Registry", "id": "mlflow-model-registry", "companyName": "MLflow project (Linux Foundation)", "companyUrl": "https://mlflow.org/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "observability" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "ML platform engineer", "MLOps lead", "Data science manager" ], "problemsSolved": [ "untracked model versions", "no lineage from model to training run", "ad hoc promotion and rollback" ], "sources": [ { "id": "s1", "url": "https://mlflow.org/docs/latest/ml/model-registry/", "title": "ML Model Registry | MLflow", "publisher": "MLflow", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents centralized model store, registered models with versions/aliases/tags/metadata, automatic version tracking with comparison and rollback, lineage to producing run or notebook, availability in OSS MLflow and managed platforms including Databricks." }, { "id": "s2", "url": "https://github.com/mlflow/mlflow", "title": "mlflow/mlflow — open source AI engineering platform", "publisher": "GitHub / MLflow project", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Public repository describing MLflow components: experiment tracking, model evaluation with 50+ metrics and LLM judges, model registry for full lifecycle management, deployment for batch and real-time scoring, prompt versioning and lineage." } ], "claims": [ { "id": "c1", "statement": "MLflow documentation states the Model Registry is a centralized model store with APIs and UI for managing the full model lifecycle, where registered models carry versions, aliases, tags and metadata, versions are tracked automatically for comparison and rollback, and lineage links each version to the run, logged model or notebook that produced it.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation covers registry mechanics; there is no governance policy engine, risk assessment or framework mapping, and Unity Catalog features described require Databricks.", "capability": "central-model-store-versioning-lineage" }, { "id": "c2", "statement": "The MLflow GitHub repository presents MLflow as an open-source AI engineering platform whose components include experiment tracking, model evaluation, a model registry supporting collaborative lifecycle management, and deployment for batch and real-time scoring.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Repository page confirms the project and component list; it does not name a vendor, and the licence text was not read on the page fetched.", "capability": "open-source-availability" } ], "limitations": [ "Licence (Apache-2.0) was not confirmed on the pages fetched this session.", "Governance features such as approvals, controls and evidence packs are outside the registry's documented scope." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Vendor field records the project rather than a company; Databricks-managed MLflow treated as a separate offering and not merged here." }, { "name": "ModelOp Center", "id": "modelop-center", "companyName": "ModelOp", "companyUrl": "https://www.modelop.com/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "ai-risk", "agent-discovery" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Model risk manager", "Chief analytics officer", "Internal audit" ], "problemsSolved": [ "incomplete model inventory", "manual model approval workflows", "control mapping across AI regulations" ], "sources": [ { "id": "s1", "url": "https://www.modelop.com/ai-governance-software/inventory", "title": "Evergreen AI Model Inventory", "publisher": "ModelOp", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes searchable registry of ML, generative AI, agents, vendor tools and embedded SaaS AI; captured artifacts, metadata, lineage; evidence including validations, controls and sign-offs; references ModelOp Center 3.3." }, { "id": "s2", "url": "https://www.modelop.com/ai-governance-software/controls", "title": "Automate AI Governance Workflows & Enforce Compliance", "publisher": "ModelOp", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States risk-tier-based requirements, automatic control mapping to EU AI Act, NIST AI RMF, OCC SR 11-7, ISO 42001, HIPAA, GDPR, lifecycle workflows from intake to retirement, automatic review triggers and blocking of non-compliant actions." } ], "claims": [ { "id": "c1", "statement": "Product page states ModelOp provides a dynamic, searchable registry of every AI solution — ML models, generative AI, agents and agentic systems, vendor tools and embedded SaaS AI — as an auditable single source of truth capturing code, prompts, agent cards, configurations, dependencies and lineage, plus evidence such as test results, validations, controls and sign-offs.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page rather than technical docs; how assets are discovered and which platforms are supported are not evidenced. Version 3.3 is referenced without release-note detail.", "capability": "ai-solution-inventory" }, { "id": "c2", "statement": "Controls page states risk-based workflows validate controls, required reviews trigger automatically and non-compliant actions can be blocked, with controls automatically mapped to internal policies and to the EU AI Act, NIST AI RMF, OCC SR 11-7, ISO 42001, HIPAA and GDPR.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Control mapping is the vendor's interpretation and does not demonstrate control effectiveness or regulatory compliance; enforcement points for blocking are unspecified.", "capability": "policy-driven-controls-and-blocking" } ], "limitations": [ "Public technical documentation and release notes were not read; capability detail rests on vendor product pages.", "Deployment options recorded from vendor positioning as enterprise software; not explicitly documented on pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "ModelOp Center is the named product; company name ModelOp. Long-standing MRM/MLOps governance vendor." }, { "name": "Modulos AI Governance Platform", "id": "modulos", "companyName": "Modulos", "companyUrl": "https://www.modulos.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI governance lead", "Compliance manager", "Risk manager" ], "problemsSolved": [ "duplicated work across AI frameworks", "untraceable governance evidence", "unquantified AI risk" ], "sources": [ { "id": "s1", "url": "https://docs.modulos.ai/platform/overview", "title": "Platform Overview | Modulos Docs", "publisher": "Modulos", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents organizations, projects, governance (frameworks, requirements, controls, evidence, reviews), risk, runtime inspection (sources, tests, schedules, results, remediation) and integrations including API tokens and Scout connectors." }, { "id": "s2", "url": "https://docs.modulos.ai/", "title": "Modulos Documentation", "publisher": "Modulos", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States AI system registration, workflows for EU AI Act, ISO/IEC 42001, NIST AI RMF and GDPR, multi-framework control mapping across 10+ frameworks, monetary risk quantification and residual-risk tracking, and an AI assistant (Scout) for evidence work." } ], "claims": [ { "id": "c1", "statement": "Documentation describes core objects (organizations, projects, governance, risk, runtime inspection, integrations) and states governance is traceable from requirement to evidence, with frameworks defining requirements, controls implementing them, evidence proving them and reviews making status changes auditable.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs describe the data model and workflow; they do not establish that produced evidence satisfies any specific regulator or auditor.", "capability": "requirement-to-evidence-traceability" }, { "id": "c2", "statement": "Documentation states structured workflows and control mapping across the EU AI Act, ISO/IEC 42001, NIST AI RMF, GDPR and 10+ frameworks simultaneously, so one control can satisfy several requirements.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Mapping breadth is asserted in docs without a published crosswalk; a mapping does not prove control effectiveness or compliance.", "capability": "multi-framework-mapping" }, { "id": "c3", "statement": "Documentation states the risk module assesses and quantifies AI risks in monetary terms using built-in methodologies and tracks residual risk across the AI portfolio.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Underlying quantification methodology is not published on the pages fetched.", "capability": "monetary-risk-quantification" } ], "limitations": [ "Swiss origin is commonly reported but was not stated on the pages fetched.", "Runtime inspection tests were not enumerated in the sections read." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Two docs pages used; marketing site not relied on for capability claims." }, { "name": "n8n", "id": "n8n", "companyName": "n8n GmbH", "companyUrl": "https://n8n.io", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "Automation engineer", "IT operations lead", "Technical founder" ], "problemsSolved": [ "building tool-using agents in workflows", "workspace access control for automations", "self-hosted automation" ], "sources": [ { "id": "s1", "url": "https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/", "title": "AI Agent node documentation - n8n Docs", "publisher": "n8n", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the AI Agent node requiring a chat model and at least one tool sub-node, the agent selecting tools to complete tasks, and the removal of the agent-type setting in version 1.82.0." }, { "id": "s2", "url": "https://docs.n8n.io/administer/manage-users-and-access/set-permissions-and-roles-rbac", "title": "Set permissions and roles (RBAC) | Administer - n8n Docs", "publisher": "n8n", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents instance roles (Owner, Admin, Member), project roles, grouping of workflows and credentials into projects, and custom roles subject to plan and deployment." } ], "claims": [ { "id": "c1", "statement": "The AI Agent node connects a chat model and at least one tool sub-node, and the agent decides which tools to call to complete a task; current versions use the Tools Agent configuration after the agent-type setting was removed in version 1.82.0.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The node documentation describes no approval, audit or permission features for tool calls.", "capability": "ai-agent-node" }, { "id": "c2", "statement": "Role-based access control operates at instance level (Owner, Admin, Member) and project level, with workflows and credentials grouped into projects and custom roles available depending on plan and deployment.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Custom role availability depends on plan and deployment, which the page does not enumerate.", "capability": "instance-and-project-rbac" }, { "id": "c3", "statement": "A per-agent audit trail, human approval gate for tool calls, or tenant isolation model is not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1", "s2" ], "limitations": "Absence of evidence; n8n log streaming documentation URLs returned errors during review.", "capability": "agent-action-audit-trail" } ], "limitations": [ "Governance is workspace-scoped, not agent-scoped: no documented approval step or audit record for individual agent tool calls.", "Documentation URLs for log streaming and AI tutorials returned 404s during review, so logging capability is unverified.", "Licensing is source-available rather than a standard OSI open-source license; this was not verified from the license file in this session." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas" ], "geographies": [ "DE", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "productType set to hybrid (source-available core plus paid cloud and enterprise tiers) rather than open_source, since the license was not verified this session." }, { "name": "Naaia", "id": "naaia", "companyName": "Naaia", "companyUrl": "https://naaia.ai/en/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Data protection officer", "Legal counsel", "AI governance lead" ], "problemsSolved": [ "EU AI Act obligation mapping", "AI operator role qualification", "centralised AI documentation" ], "sources": [ { "id": "s1", "url": "https://naaia.ai/en/eu-ai-act/", "title": "EU AI Act Compliance – Automate Your Obligations | Naaia", "publisher": "Naaia", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes operationalising EU AI Act compliance via qualification of operator status and risk level plus an automated operational action plan; explains the Act's risk tiers and GPAI category." }, { "id": "s2", "url": "https://naaia.ai/en/", "title": "AI Governance & Compliance Platform | Naaia", "publisher": "Naaia", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Naaia as Europe's first AI Management System based on ISO/IEC 42001, centralized registry for projects, systems, models and components, and states UGAP and CANUT catalogue listings." } ], "claims": [ { "id": "c1", "statement": "Product page states Naaia provides integrated qualification of operator status and risk level under the EU AI Act with an automated operational action plan covering the Act's obligations.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page. Automated qualification is a vendor tool output and does not establish legal compliance with the EU AI Act; no documentation of the underlying rules was found.", "capability": "eu-ai-act-qualification-and-action-plan" }, { "id": "c2", "statement": "Homepage positions the product as a European AI management system built on ISO/IEC 42001, with a centralised registry managing projects, systems, models and components as a single source of truth, and states listing in the UGAP and CANUT public-procurement catalogues.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Being built 'on' ISO/IEC 42001 is not certification; catalogue listings were not independently checked.", "capability": "aims-and-central-registry" } ], "limitations": [ "No technical documentation, API reference or release notes were located.", "'First European AIMS' is a vendor superlative and was not verified." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Geography EU recorded because the vendor pages explicitly scope the product to Europe/EU AI Act; specific HQ city not stated." }, { "name": "NannyML", "id": "nannyml", "companyName": "NannyML", "companyUrl": "https://www.nannyml.com/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "observability" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Data scientist", "ML engineer", "Model validator" ], "problemsSolved": [ "performance blind spots without labels", "drift alerts without business impact", "post-deployment model decay detection" ], "sources": [ { "id": "s1", "url": "https://github.com/NannyML/NannyML", "title": "NannyML/nannyml — post-deployment data science in python", "publisher": "GitHub / NannyML", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents performance estimation without targets, CBPE and DLE, drift detection linked to performance change, tabular classification and regression support, realised performance tracking, PCA-based multivariate drift and univariate statistical tests; version 0.12.1 referenced." }, { "id": "s2", "url": "https://nannyml.readthedocs.io/en/stable/", "title": "Welcome to NannyML's documentation!", "publisher": "NannyML", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents performance estimation and calculation, CBPE/DLE for classification and regression, business value estimation, confusion matrix elements, custom metrics and univariate/multivariate drift monitoring; version 0.13.1." } ], "claims": [ { "id": "c1", "statement": "The NannyML repository states the library estimates post-deployment model performance without access to targets, detects data drift, links drift alerts to changes in model performance, supports tabular classification and regression, and implements confidence-based performance estimation (CBPE) and direct loss estimation (DLE), including metrics such as ROC AUC and RMSE.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documented scope is tabular ML; the repository documents no registry, approvals or lifecycle governance features.", "capability": "performance-estimation-without-targets" }, { "id": "c2", "statement": "NannyML's stable documentation (version 0.13.1) documents performance estimation and calculation, business-value estimation, confusion-matrix elements, custom binary classification metrics, and univariate and multivariate data drift monitoring.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Docs describe methods and metrics but not enterprise features, and no vendor company details are given.", "capability": "documented-monitoring-methods" }, { "id": "c3", "statement": "Multiple repository paths exist for the project (NannyML/NannyML plus forks such as nnansters/nannyml), so the canonical source should be confirmed before adoption.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Only establishes that several repository names surface publicly; ownership history was not investigated.", "capability": "repository-fragmentation" } ], "limitations": [ "Licence identifier not confirmed on the pages fetched.", "Commercial offering or company status was not evidenced; only the library was verified." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "issue": "Version mismatch", "detail": "GitHub page references version 0.12.1 while the stable docs show 0.13.1.", "sourceIds": [ "s1", "s2" ] } ], "editorialNotes": "Included for lifecycle/drift monitoring under model-governance; no registry or approval features." }, { "name": "Netskope One AI Command Center", "id": "netskope-ai-command-center", "companyName": "Netskope", "companyUrl": "https://www.netskope.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "data-governance", "agent-security" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Network security architect", "Data protection lead" ], "problemsSolved": [ "shadow AI application discovery", "AI asset to data-store mapping", "correlated AI risk response" ], "sources": [ { "id": "s1", "url": "https://www.netskope.com/press-releases/netskope-unveils-ai-command-center-delivering-comprehensive-ai-discovery-and-correlated-risk-intelligence-with-fully-coordinated-agentic-response", "title": "Netskope Unveils AI Command Center", "publisher": "Netskope", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-06-02", "supports": "Announces AI Command Center discovering AI assets across corporate/personal, managed/shadow and cloud/on-premises, mapping them to identities, data stores and tools." }, { "id": "s2", "url": "https://www.netskope.com/solutions/netskope-one-ai-security", "title": "Securing AI with Netskope One", "publisher": "Netskope", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the suite gives unified visibility across users, apps, MCP servers and autonomous agents from pre-deployment to runtime." } ], "claims": [ { "id": "c1", "statement": "AI Command Center discovers AI assets whether corporate or personal, managed or shadow, cloud or on-premises, and maps them to the identities, data stores and tools they connect to, surfacing attack paths.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Press release announced 2 June 2026; it does not document the detection mechanism per asset class or list supported agent platforms.", "capability": "ai-asset-discovery-and-mapping" }, { "id": "c2", "statement": "Netskope One AI Security provides a unified view of AI usage across users, apps and MCP servers, covering generative AI tools, private AI applications and autonomous agents operating via APIs and MCP.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Solution page; per-capability availability and enforcement depth for agent traffic are not specified.", "capability": "unified-ai-visibility" } ], "limitations": [ "No product documentation reviewed for AI Command Center", "Detection technique per asset class is not publicly specified" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Statistics quoted in the press release are Netskope Threat Labs telemetry and were not treated as verified market data." }, { "name": "NeuralTrust (TrustGate)", "id": "neuraltrust", "companyName": "NeuralTrust", "companyUrl": "https://neuraltrust.ai/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "runtime-controls", "agent-discovery" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AI platform owner", "Compliance lead" ], "problemsSolved": [ "per-tool RBAC for agents", "identity propagation across agent hops", "runtime monitoring of agent behaviour" ], "sources": [ { "id": "s1", "url": "https://neuraltrust.ai/", "title": "NeuralTrust | The Platform for AI and Agent Security", "publisher": "NeuralTrust", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the platform controls agent behaviour and access from reasoning to action execution, protects models, tools, MCP servers, skills and data from one runtime layer and maintains a real-time inventory of AI operations." }, { "id": "s2", "url": "https://neuraltrust.ai/ai-gateway", "title": "AI Gateway for Enterprise - TrustGate", "publisher": "NeuralTrust", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes TrustGate as an agent gateway with per-user tool access examples, per-agent and per-tool RBAC, end-user identity forwarding, cryptographic audit trails and SaaS, hybrid or air-gapped deployment." } ], "claims": [ { "id": "c1", "statement": "The TrustGate page states end-user identity is forwarded through every hop with per-agent and per-tool RBAC, showing examples of allowed reads and a denied payments API, and cryptographic audit trails recording every tool call.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Policy language, evaluation semantics and integration prerequisites are not documented on the page.", "capability": "gateway-rbac-and-identity" }, { "id": "c2", "statement": "NeuralTrust describes a single runtime layer protecting what agents access and execute across models, tools, MCP servers, skills and data, with monitoring of agent behaviour to stop attacks and unusual activity.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Detection techniques and coverage are not described in reference documentation.", "capability": "runtime-protection-scope" }, { "id": "c3", "statement": "TrustGate is offered as SaaS, hybrid cloud and on-premises or air-gapped deployment, with self-service sign-up available.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Feature differences between deployment modes are not stated.", "capability": "deployment-options" } ], "limitations": [ "Evidence limited to vendor product pages", "Open-source status of TrustGate not stated on fetched pages" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid", "self_hosted" ], "geographies": [ "ES", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "NeuralTrust also markets a 'Generative Application Firewall'; treated as part of the same platform rather than a separate product." }, { "name": "Noma Security Platform", "id": "noma-security", "companyName": "Noma Security", "companyUrl": "https://noma.security/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "agent-discovery", "evaluation", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AI security lead", "Head of platform security" ], "problemsSolved": [ "agent and MCP server discovery", "runtime action blocking and human routing", "pre-production agent red teaming" ], "sources": [ { "id": "s1", "url": "https://noma.security/platform", "title": "Govern and Secure AI Agents Everywhere They Run", "publisher": "Noma Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes discovery, posture, access control, adversarial testing and runtime detection across endpoint, SaaS and homegrown agents, and an AI Control Plane spanning discovery, access control and runtime enforcement." }, { "id": "s2", "url": "https://www.noma.security/platform/homegrown-agents", "title": "AI Application Security for Bedrock, Azure & Databricks", "publisher": "Noma Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Noma builds a continuous inventory of agents, models, MCP servers and skills, maps blast radius, evaluates each action in context to alert, block, mask data or route to a human, and red teams agents pre-production." } ], "claims": [ { "id": "c1", "statement": "Noma states it discovers every agent, MCP server, skill and model across endpoint, SaaS and homegrown environments and correlates each agent's connections, permissions and data access to surface toxic combinations.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Discovery mechanisms and platform coverage limits are not documented in reference material.", "capability": "agent-inventory-and-posture" }, { "id": "c2", "statement": "Runtime protection is described as evaluating each action in context and, based on policy, alerting, blocking, masking data or routing to a human, catching prompt injection, data exfiltration and scope violations.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Human-approval routing is asserted on a solution page without workflow documentation.", "capability": "runtime-enforcement-options" }, { "id": "c3", "statement": "The platform is described as including adversarial testing and red teaming that compound techniques into multi-turn campaigns before agents reach production.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Testing methodology, coverage and reporting formats are not documented.", "capability": "adversarial-testing" } ], "limitations": [ "No public reference documentation fetched", "Availability stage and deployment options not stated" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "IL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Two domains (noma.security and www.noma.security) serve the same product; kept as one entry." }, { "name": "Nudge Security AI Agent Discovery", "id": "nudge-security-ai-agent-discovery", "companyName": "Nudge Security", "companyUrl": "https://www.nudgesecurity.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "identity-access", "ai-risk" ], "productType": "commercial", "availability": "preview", "buyerRoles": [ "CISO", "IT security manager", "SaaS governance lead" ], "problemsSolved": [ "employee-created agent discovery", "MCP connection visibility", "orphaned and over-permissioned agent cleanup" ], "sources": [ { "id": "s1", "url": "https://www.nudgesecurity.com/try-nudge/ai-agent-discovery", "title": "AI Agent Discovery & Governance", "publisher": "Nudge Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Research preview page listing supported agent platforms, creator attribution, permission chains, OAuth/API-token-based agent connections, MCP visibility and risk findings." }, { "id": "s2", "url": "https://www.nudgesecurity.com/post/ai-discovery-methods-compared", "title": "Shadow AI discovery: Detection methods compared", "publisher": "Nudge Security", "type": "other", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Explains email-based, browser-based, network and SaaS-API discovery methods and their respective blind spots." } ], "claims": [ { "id": "c1", "statement": "Discovers AI agents as employees create them, including agents deployed without approval, showing who created each agent, what it is connected to, its permissions and capabilities, and what data it can access, across Copilot Studio, Agentforce, Gemini, OpenAI, n8n, ServiceNow, Databricks, Workato and Tines.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Explicitly labelled a research preview; the page does not document the technical connector used per platform.", "capability": "creator-attributed-agent-inventory" }, { "id": "c2", "statement": "Surfaces AI connections created through OAuth grants and API tokens and answers what MCP activity exists in an environment, flagging MCP connections with no authentication.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documentation reviewed on how MCP activity is observed or how complete that view is.", "capability": "oauth-and-mcp-visibility" }, { "id": "c3", "statement": "Nudge describes four shadow AI detection methods - email-based discovery, browser-based discovery, network monitoring, and SaaS API connections used by SSPM tools - and states each has blind spots so comprehensive solutions combine several.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "This is Nudge's own framing of the market; it does not state which combination Nudge itself implements for every platform.", "capability": "discovery-method-taxonomy" } ], "limitations": [ "Research preview status means feature set and platform coverage may change", "Per-platform detection mechanism is not documented publicly" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included because the vendor names specific agent platforms and attribution fields; availability set to preview per the page's own 'Research Preview' label." }, { "name": "NVIDIA NeMo Guardrails", "id": "nemo-guardrails", "companyName": "NVIDIA", "companyUrl": "https://www.nvidia.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "open_source", "availability": "preview", "buyerRoles": [ "AI engineer", "ML platform owner", "Application security engineer" ], "problemsSolved": [ "programmable input and output rails", "jailbreak detection", "constraining tool/action execution in LLM apps" ], "sources": [ { "id": "s1", "url": "https://github.com/NVIDIA/NeMo-Guardrails/", "title": "NVIDIA/NeMo-Guardrails", "publisher": "NVIDIA (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes an open-source Apache-2.0 toolkit for programmable guardrails with jailbreak and prompt injection protections and execution rails on custom actions; states main tracks beta 0.12.0 and production deployment is not recommended." }, { "id": "s2", "url": "https://docs.nvidia.com/nemo/guardrails/home", "title": "NVIDIA NeMo Guardrails Library Developer Guide", "publisher": "NVIDIA", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the library intercepts inputs and outputs, applies configurable safety checks, blocks or modifies content per policy, and lists jailbreak detection, PII handling and agentic security in the guardrail catalogue." } ], "claims": [ { "id": "c1", "statement": "NeMo Guardrails intercepts inputs and outputs of LLM applications, applies configurable safety checks and blocks or modifies content based on defined policies, configured with YAML files, Colang flows and custom actions.", "support": "documented", "sourceIds": [ "s2", "s1" ], "limitations": "Rail effectiveness depends on operator configuration and chosen models; no benchmark is provided on the fetched pages.", "capability": "programmable-rails" }, { "id": "c2", "statement": "The toolkit provides mechanisms against jailbreaks and prompt injections, includes jailbreak detection, PII handling and agentic security entries in its guardrail catalogue, and offers execution rails applied to the input and output of custom actions or tools.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Agentic security catalogue entry is listed without detailed threat coverage on the fetched pages.", "capability": "injection-and-agentic-checks" }, { "id": "c3", "statement": "The repository states the main branch tracks beta version 0.12.0 and that NVIDIA does not currently recommend deploying this beta version in a production setting.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The NVIDIA documentation site presents the library as available without repeating the production caveat.", "capability": "beta-status" } ], "limitations": [ "Beta status per repository", "Requires engineering effort to author rails" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "self_hosted" ], "geographies": [ "global", "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ "Repository labels the release beta and not production-ready while the NVIDIA developer guide presents it as an available library" ], "editorialNotes": "Two repositories exist (NVIDIA/NeMo-Guardrails and NVIDIA-NeMo/Guardrails); treated as one product with documentation consolidated on docs.nvidia.com. Also surfaced in the open-source pass; runtime-controls kept as primary." }, { "name": "Oasis Agentic Access Management (AAM)", "id": "oasis-agentic-access-management", "companyName": "Oasis Security", "companyUrl": "https://oasis.security", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "IAM manager", "Security architect" ], "problemsSolved": [ "standing privileges for AI agents", "auditing what an agent did and why", "provisioning just-in-time agent access" ], "sources": [ { "id": "s1", "url": "https://www.oasis.security/blog/introducing-oasis-agentic-access-management", "title": "Introducing Oasis Agentic Access Management (AAM)", "publisher": "Oasis Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-05-01", "supports": "Describes intent-aware, policy-driven identity infrastructure converting agent interactions into short-lived least-privilege sessions with prompt-level audit." }, { "id": "s2", "url": "https://www.prnewswire.com/news-releases/oasis-security-launches-agentic-access-management-the-first-identity-solution-built-for-ai-agents-302619375.html", "title": "Oasis Security Launches Agentic Access Management", "publisher": "PR Newswire", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-11-19", "supports": "Launch release detailing intent inference, deterministic policy enforcement, JIT session identities and lifecycle governance." }, { "id": "s3", "url": "https://oasis.security/", "title": "Non Human Identity Management Platform | OASIS Security", "publisher": "Oasis Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States agent identity provisioning, policy enforcement at scale and NHI coverage across IaaS, SaaS, PaaS and on-prem." } ], "claims": [ { "id": "c1", "statement": "Oasis AAM provisions ephemeral, least-privilege session identities scoped to a single action, eliminating standing secrets, with audit trails covering person, prompt, policy, actions and teardown.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Neither source is reference documentation; supported target systems for provisioning are not enumerated.", "capability": "jit-session-identities" }, { "id": "c2", "statement": "The product infers agent intent from prompts, tool calls and action plans and evaluates it deterministically against enterprise policy, escalating to a human when privilege boundaries are crossed.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Accuracy of LLM-driven intent inference is not independently validated.", "capability": "intent-based-policy" }, { "id": "c3", "statement": "AAM extends Oasis's existing non-human identity management platform, which covers NHIs across IaaS, SaaS, PaaS and on-prem environments.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "Home page does not document audit or credential issuance specifics.", "capability": "nhi-platform-base" } ], "limitations": [ "No public documentation portal was fetched; claims rest on vendor product and press pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Launch release dated Nov 2025; product blog May 2026. Availability treated as GA based on the launch release language." }, { "name": "Obsidian Security Shadow AI", "id": "obsidian-shadow-ai", "companyName": "Obsidian Security", "companyUrl": "https://www.obsidiansecurity.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "agent-security", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "SaaS security lead", "Security operations manager" ], "problemsSolved": [ "shadow AI tool inventory", "unapproved agent detection", "prompt data exfiltration control" ], "sources": [ { "id": "s1", "url": "https://www.obsidiansecurity.com/shadow-ai-security", "title": "Shadow AI Security: Discover, Govern, and Control Every AI Tool", "publisher": "Obsidian Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Obsidian combines browser-level discovery, API integration scanning and agent monitoring, maps MCP connections, and enforces guardrails via a browser extension." }, { "id": "s2", "url": "https://www.obsidiansecurity.com/blog/ai-agents-security-tools-cant-see", "title": "The AI Agents Your Security Tools Can't See", "publisher": "Obsidian Security", "type": "other", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Vendor blog arguing API-based agent inventory is incomplete, with customer-base agent counts rising from under 500 in late 2024 to about 95,000 by February 2026." } ], "claims": [ { "id": "c1", "statement": "Combines browser-level discovery, API integration scanning, and agent monitoring into a single continuously updated inventory of AI tools, AI features inside SaaS apps, and agents.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product marketing page; does not enumerate which SaaS platforms are covered by API scanning or which browsers the extension supports.", "capability": "multi-signal-ai-discovery" }, { "id": "c2", "statement": "Vendor states that API-only agent inventories miss agents that were not created through official channels, and reports agent counts across its customer base growing to roughly 95,000 by February 2026.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Self-reported telemetry from Obsidian deployments; not independently verified and not a capability guarantee.", "capability": "api-inventory-gap" }, { "id": "c3", "statement": "A browser extension monitors and can block sensitive data entered into third-party AI tools in real time.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documentation reviewed on detection accuracy, supported browsers, or bypass behavior for native desktop AI apps.", "capability": "prompt-dlp-blocking" } ], "limitations": [ "No public documentation reviewed listing supported agent platforms or extension-managed browsers", "Statistics are vendor telemetry, not audited" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Distinct from Valence and AppOmni by explicitly pairing browser-extension discovery with SaaS API scanning. Funding coverage seen in search results was not used as evidence." }, { "name": "Okta for AI Agents", "id": "okta-for-ai-agents", "companyName": "Okta", "companyUrl": "https://www.okta.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "commercial", "availability": "unknown", "buyerRoles": [ "CISO", "IAM manager", "Identity governance lead" ], "problemsSolved": [ "shadow AI agent discovery", "agent registration and ownership", "replacing long-lived agent tokens" ], "sources": [ { "id": "s1", "url": "https://www.okta.com/products/govern-ai-agent-identity/", "title": "Okta for AI Agents | Govern Agentic Identity", "publisher": "Okta", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-07-23", "supports": "Describes discovery of known and shadow agents, registration in Universal Directory, owner assignment, short-lived credentials, secret vaulting and least-privilege policy." }, { "id": "s2", "url": "https://help.okta.com/oie/en-us/content/topics/ai-agents/ai-agent-inventory.htm", "title": "AI agent discovery in ISPM - Okta Documentation", "publisher": "Okta", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the Discovered agents page, OAuth-grant based shadow agent detection via the Secure Access Monitor plugin, Agentforce coverage, and SKU exclusions." } ], "claims": [ { "id": "c1", "statement": "AI agents and MCP servers can be registered as first-class identities in Okta Universal Directory with an assigned human owner.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page does not state which agent platforms can be registered automatically versus manually.", "capability": "agent-registration-directory" }, { "id": "c2", "statement": "Okta Identity Security Posture Management discovers AI agents in managed and unmanaged apps, shows owners, permissions and impacted resources, and can register discovered agents in Okta.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documentation states discovery currently covers agents built with Salesforce Agentforce; ISPM is excluded from the Okta for AI Agents Core SKU.", "capability": "ispm-agent-discovery" }, { "id": "c3", "statement": "The product provides agents with temporary, short-lived credentials and vaults and rotates secrets and API keys.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documented token lifetime, rotation interval or supported target systems on the fetched page.", "capability": "short-lived-agent-credentials" } ], "limitations": [ "Availability stage (GA vs preview) of individual capabilities was not stated on fetched pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept separate from Auth0 (same parent, Okta Inc.) because Auth0's agent auth targets application developers rather than workforce identity admins." }, { "name": "OneTrust AI Governance", "id": "onetrust-ai-governance", "companyName": "OneTrust", "companyUrl": "https://www.onetrust.com/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "agent-discovery", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Privacy officer", "AI governance lead" ], "problemsSolved": [ "shadow AI usage", "inconsistent AI risk tiering", "evidence for AI audits" ], "sources": [ { "id": "s1", "url": "https://www.onetrust.com/products/ai-governance/", "title": "AI Governance — OneTrust", "publisher": "OneTrust", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes OneTrust AI Governance: central inventory of models, datasets, agents and vendors, use-case intake and approval workflows, risk tiering automation, impact assessments and framework-based risk identification." }, { "id": "s2", "url": "https://www.onetrust.com/solutions/ai-governance/", "title": "AI Governance Software | Solutions", "publisher": "OneTrust", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims continuous AI discovery and inventory, automated intake, risk scoring and approvals, policy-driven controls, runtime enforcement, attestations and audit-ready evidence; notes Visionary placement in 2026 Gartner MQ for AI Governance Platforms." } ], "claims": [ { "id": "c1", "statement": "Product page states a central inventory of models, datasets, agents and vendors with ownership and lifecycle status, configurable AI use-case intake and approval workflows, automated risk tiering by use case/system/component, and impact assessments.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product marketing page; no technical documentation was fetched, so workflow configurability and discovery coverage are unverified.", "capability": "ai-inventory-intake-approvals" }, { "id": "c2", "statement": "Solutions page claims continuous discovery of AI across environments, policy evaluation across models, data, agents and vendors, real-time guardrails and runtime enforcement, and generation of audit-ready evidence with attestations.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing claims only. Runtime enforcement scope, supported platforms and what 'audit-ready' means for any specific framework are not evidenced.", "capability": "runtime-controls-and-evidence" } ], "limitations": [ "This is an AI module within a broader GRC/privacy suite; licence boundaries versus other OneTrust products were not evidenced.", "No product documentation (support portal requires login) was consulted." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included because OneTrust publishes a distinct AI Governance product page, satisfying the 'named AI module' rule." }, { "name": "OpenFGA", "id": "openfga", "companyName": "OpenFGA project", "companyUrl": "https://openfga.dev/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "identity-access" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Application security architect", "IAM engineer", "Platform engineer" ], "problemsSolved": [ "scoped delegation to agents", "MCP tool-level permission modelling", "revoking agent access independently" ], "sources": [ { "id": "s1", "url": "https://openfga.dev/docs/use-cases/ai-agent-authorization", "title": "AI Agent Authorization with OpenFGA", "publisher": "OpenFGA project", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States agents are first-class principals, permissions are delegated not copied, agents can be revoked independently, and lists task-based, RAG and MCP server authorization patterns." }, { "id": "s2", "url": "https://openfga.dev/docs/modeling/agents", "title": "Authorization for Agents - OpenFGA", "publisher": "OpenFGA project", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents zero-permission agents with narrowly scoped per-task grants, expiration, turn limits and agent binding, plus controlling which tools each user can access on an MCP server." }, { "id": "s3", "url": "https://github.com/openfga/openfga", "title": "openfga/openfga", "publisher": "OpenFGA (GitHub)", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms Apache-2.0 license and a high-performance fine-grained authorization engine with HTTP and gRPC APIs." } ], "claims": [ { "id": "c1", "statement": "OpenFGA documentation models agents as first-class principals that appear in authorization tuples, receive delegated rather than copied permissions, and can be revoked independently of the user they serve.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Modelling guidance is not an enforcement mechanism; interception must be built by the adopter.", "capability": "agents-as-principals" }, { "id": "c2", "statement": "Agents can start with zero permissions and receive narrowly scoped grants per task, with optional expiration, turn limits and agent binding, and MCP tool access can be controlled per user by role, group and temporal grant.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Documented as modelling patterns; no reference implementation of an MCP interceptor is provided on these pages.", "capability": "scoped-task-grants" }, { "id": "c3", "statement": "OpenFGA is an Apache-2.0 licensed, high-performance authorization engine inspired by Google Zanzibar with HTTP and gRPC APIs.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Repository evidence does not establish a vendor support commitment.", "capability": "engine-and-license" } ], "limitations": [ "Not a gateway or interception point", "No prompt or content inspection" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "OpenFGA originated at Okta/Auth0 and is a CNCF project; no ownership statement was found on the fetched pages, so companyName is recorded as the project." }, { "name": "OpenLLMetry", "id": "openllmetry", "companyName": "Traceloop", "companyUrl": "https://www.traceloop.com", "primaryCategoryId": "observability", "secondaryCategoryIds": [], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Platform engineer", "Observability lead" ], "problemsSolved": [ "standards-based instrumentation of LLM apps", "exporting GenAI traces to existing observability stacks", "vendor-neutral tracing" ], "sources": [ { "id": "s1", "url": "https://www.traceloop.com/docs/openllmetry/introduction", "title": "What is OpenLLMetry?", "publisher": "Traceloop", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes OpenLLMetry as OpenTelemetry-based non-intrusive tracing for LLM apps, Python/JS/Go SDKs, export to Traceloop or an existing observability stack, and documentation sections for workflows, tasks, agents and tools plus privacy." }, { "id": "s2", "url": "https://github.com/traceloop/openllmetry", "title": "traceloop/openllmetry", "publisher": "GitHub / Traceloop", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States open-source observability for GenAI or LLM applications based on OpenTelemetry, Apache 2.0 license, traceloop repository owner, OpenLLMetry-JS companion, and that the SDK no longer logs or collects telemetry." } ], "claims": [ { "id": "c1", "statement": "OpenLLMetry is a set of extensions built on OpenTelemetry that provides observability over an LLM application with non-intrusive tracing, with Python, JavaScript/TypeScript and Go SDKs and documentation for annotating workflows, tasks, agents and tools.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The introduction does not state which attributes are captured per span, nor cost, token or evaluation support.", "capability": "otel-based-instrumentation" }, { "id": "c2", "statement": "Traces can be exported to Traceloop or to an existing observability stack, and the repository states the project is Apache 2.0 licensed under the traceloop organization with a companion OpenLLMetry-JS project.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "No governance features (audit logs, RBAC, retention, redaction, certifications) are stated in either source.", "capability": "export-targets-and-license" }, { "id": "c3", "statement": "The repository states the project no longer logs or collects any telemetry in the SDK or instrumentations, and that any previously collected data was anonymous with no personally identifiable information.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Statement covers the library's own telemetry, not customer data handling by the Traceloop hosted service.", "capability": "sdk-telemetry-policy" } ], "limitations": [ "Instrumentation library only: no storage, retention, access control or investigation UI of its own.", "Adherence to OpenTelemetry GenAI semantic conventions is not documented on the pages reviewed." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "IL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "OpenTelemetry", "type": "native", "sourceIds": [ "s1", "s2" ], "limitations": "Built on OpenTelemetry; conformance with GenAI semantic conventions was not verified on the pages fetched." } ], "conflicts": [], "editorialNotes": "Recorded as the library rather than the Traceloop hosted platform, which was not verified in this session." }, { "name": "OpenSSF Model Signing (OMS) / model-transparency", "id": "openssf-model-signing", "companyName": "Open Source Security Foundation", "companyUrl": "https://openssf.org", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "data-governance", "agent-security" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "ML platform engineer", "Supply chain security lead", "CISO" ], "problemsSolved": [ "model integrity verification", "ML supply chain provenance", "signing large model artifacts" ], "sources": [ { "id": "s1", "url": "https://openssf.org/projects/model-signing/", "title": "OpenSSF Model Signing (OMS)", "publisher": "OpenSSF", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the library and CLI, format and size support, PKI options and verification points." }, { "id": "s2", "url": "https://github.com/sigstore/model-transparency", "title": "sigstore/model-transparency", "publisher": "Sigstore / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the signing mechanism, Sigstore bundle/DSSE/in-toto structure and PKCS#11 support." }, { "id": "s3", "url": "https://api.github.com/repos/sigstore/model-transparency", "title": "GitHub REST API repository record", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Reports Apache-2.0 license, archived=false, last push 2026-09-04." } ], "claims": [ { "id": "c1", "statement": "OpenSSF states the project provides a library and CLI for signing and verification of ML models, supporting any model format and size and several PKI options including Sigstore, self-signed certificates and public/private key pairs, with signatures checkable whenever a model is uploaded, deployed or reused.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Signatures attest to artifact integrity and signer identity; they say nothing about model quality, safety or training-data legality.", "capability": "signing-library-and-cli" }, { "id": "c2", "statement": "The sigstore/model-transparency repository states the signing process produces a Sigstore bundle stored as JSON containing a DSSE envelope with an in-toto statement whose subjects are file path and digest pairs, and supports PKCS#11 devices.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Repository documents the mechanism; adoption across model hubs is not evidenced here.", "capability": "implementation-details" }, { "id": "c3", "statement": "GitHub metadata for sigstore/model-transparency records Apache-2.0 licensing, a non-archived repository and commits in September 2026.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "The OpenSSF project page itself does not state a license.", "capability": "license-and-activity" } ], "limitations": [ "Specification and implementation live in different organisations (OpenSSF spec, sigstore repo)", "Signature verification is only useful with a policy that enforces it" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "self_hosted" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included because the specification has a usable implementation; the spec alone would have been held per brief instructions." }, { "name": "Operant Semantic Firewall", "id": "operant-semantic-firewall", "companyName": "Operant AI", "companyUrl": "https://www.operant.ai/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "agent-security" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Cloud security architect", "Platform security lead" ], "problemsSolved": [ "pre-execution blocking of unsafe agent actions", "least-privilege MCP tool usage", "runtime data redaction" ], "sources": [ { "id": "s1", "url": "https://www.operant.ai/platform/semantic-firewall", "title": "Operant Semantic Firewall | Govern AI Agents in Real Time", "publisher": "Operant AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes inline allow/block/redact decisions on every agent action, pre-execution inspection of generated commands, and blocking of exfiltration, bulk dumps and credential access." }, { "id": "s2", "url": "https://www.operant.ai/solutions/mcp-gateway", "title": "MCP Gateway - Operant AI", "publisher": "Operant AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes least-privilege execution controls and access permissions for MCP tool usage, trust zones with real-time blocking of untrusted servers and tools, and detection of prompt injection, jailbreaks and tool poisoning." } ], "claims": [ { "id": "c1", "statement": "The product page states the Semantic Firewall reads the meaning of every prompt, plan, tool call, command and data payload before execution and enforces an allow, block or redact decision inline.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No reference documentation, benchmark or latency figure was available on the fetched pages.", "capability": "inline-allow-block-redact" }, { "id": "c2", "statement": "Operant states it separates legitimate work from injected instructions, blocks privilege escalation and shell breakout attempts pre-execution, and stops bulk reads, credential access and unauthorised sharing across MCP servers, plugins and sub-agents.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Detection methodology and false-positive handling are not described.", "capability": "injection-and-exfiltration-blocking" }, { "id": "c3", "statement": "The MCP Gateway page states it enforces least-privilege execution controls and per-tool access permissions, with real-time blocking of untrusted servers and tools and detection of prompt injection, jailbreaks and tool poisoning.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Both sources are marketing pages; deployment prerequisites are not documented.", "capability": "mcp-gateway-controls" } ], "limitations": [ "No official reference documentation fetched", "Availability stage not explicitly stated by the vendor" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "hybrid" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Placed under runtime controls because the evidenced core function is pre-execution allow/block/redact enforcement; agent-security recorded as secondary." }, { "name": "Opik", "id": "opik", "companyName": "Comet", "companyUrl": "https://www.comet.com", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI engineer", "ML platform lead" ], "problemsSolved": [ "agent and LLM call tracing", "cost and error monitoring", "online scoring of production traces", "turning failures into regression test cases" ], "sources": [ { "id": "s1", "url": "https://www.comet.com/docs/opik/", "title": "Opik Documentation", "publisher": "Comet", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Opik records every LLM call, tool invocation and agent step, provides a complete log of requests and responses, dashboards for cost, latency, error rates and feedback scores, online evaluation rules scoring incoming traces, prompt versioning and turning traces into test cases." }, { "id": "s2", "url": "https://github.com/comet-ml/opik", "title": "comet-ml/opik", "publisher": "GitHub / Comet", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Apache-2.0 license, built by Comet, open source covering the full platform including server backend and web application, free self-hosting with no data leaving the environment, and Opik Guardrails features." } ], "claims": [ { "id": "c1", "statement": "Opik records every LLM call, tool invocation and agent step, providing a complete log of every request and response, with dashboards for cost, latency, error rates and feedback scores and token cost tracking.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs do not use span terminology or describe multi-agent handoff modelling; retention behaviour is unstated.", "capability": "trace-capture" }, { "id": "c2", "statement": "Online evaluation rules automatically score incoming traces using LLM-as-a-judge and heuristic metrics, and a problematic trace can be turned into a test case added via UI or SDK and run as a suite.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented approval workflow or gate tied to those scores.", "capability": "online-evaluation-and-tests" }, { "id": "c3", "statement": "The repository states the full platform including server backend, web application, tracing, datasets, experiments, evaluations, prompt management, online evaluation and agent optimization components is Apache-2.0 licensed and free to self-host with no data leaving the environment, and includes Opik Guardrails.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No RBAC, SSO, SCIM, audit log, retention or PII-masking features are stated in the repository README.", "capability": "self-hosting-and-license" } ], "limitations": [ "Self-hosting shifts data control to the buyer but no in-product access control, audit trail or retention policy is documented.", "Guardrails are mentioned in the repository without documented policy definitions or enforcement points." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Hybrid: Apache-2.0 self-hostable platform plus Comet-hosted cloud." }, { "name": "ORCAA Algorithmic Audit", "id": "orcaa-algorithmic-audit", "companyName": "ORCAA (O'Neil Risk Consulting & Algorithmic Auditing)", "companyUrl": "https://orcaarisk.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk", "evaluation" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "General Counsel", "Head of HR compliance", "Chief Risk Officer" ], "problemsSolved": [ "algorithmic bias audit", "regulatory bias testing", "AI governance advice" ], "sources": [ { "id": "s1", "url": "https://orcaarisk.com/", "title": "ORCAA", "publisher": "ORCAA", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists Algorithmic Audit, quantitative bias testing pilot, AI governance and risk management consultation and cockpit design with stated deliverables." }, { "id": "s2", "url": "https://orcaarisk.com/about", "title": "About ORCAA", "publisher": "ORCAA", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the firm's purpose of defining accountability for algorithms and developing frameworks and governance approaches." } ], "claims": [ { "id": "c1", "statement": "ORCAA describes an Algorithmic Audit producing an audit report using its Ethical Matrix framework, and quantitative bias testing on live or test data delivering bias audit reports such as those for NYC Local Law 144.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Service page; no sample report, statistical methodology detail, or evidence of regulator acceptance of specific reports.", "capability": "audit-and-bias-testing" }, { "id": "c2", "statement": "ORCAA states it provides AI governance and risk management consultation covering governance policies, processes and structures and vendor diligence reports, and designs monitoring 'cockpits' of risk metrics.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Descriptive mission and service copy; no methodology documentation or independent validation.", "capability": "governance-consulting" } ], "limitations": [ "Small-firm capacity and geographic reach not documented", "Patent-pending analysis platform is described but not documented technically" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "US" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Bias-audit demand is largely US-jurisdiction driven (NYC LL144); geography recorded conservatively as US." }, { "name": "Orkes Conductor", "id": "orkes-conductor", "companyName": "Orkes", "companyUrl": "https://orkes.io", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "Platform engineering lead", "Backend architect", "AI engineering manager" ], "problemsSolved": [ "durable agent execution", "agent versioning and reuse", "visibility into agent steps" ], "sources": [ { "id": "s1", "url": "https://orkes.io/content/devguide/ai/conductor-agents", "title": "Conductor Agents | Orkes Docs", "publisher": "Orkes", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-05-14", "supports": "Documents authoring agents in code or from supported frameworks, compilation into workflow definitions, plan/deploy/serve/run lifecycle, versioned registration with optional version pinning, AGENT task invocation, and UI/API visibility of every LLM call, tool invocation, wait, retry and branch, plus composition with human approval and cancellation." }, { "id": "s2", "url": "https://orkes.io/content/access-control-and-security/users-and-groups", "title": "Managing Users and Groups | Orkes Docs", "publisher": "Orkes", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents users, groups, roles, resource permissions and inherited permissions across workflows, tasks, secrets, integrations, prompts and gateway services, noting some features are admin-only." } ], "claims": [ { "id": "c1", "statement": "Agents authored in Python, Java, TypeScript or C#, or imported from frameworks such as OpenAI Agents, Google ADK and LangGraph, compile into ordinary workflow definitions and are registered on the server as reusable, versioned Conductor Agents, with an option to pin a version when invoking them.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation does not describe approval gates on version promotion or a change-audit record.", "capability": "agent-compilation-and-versioning" }, { "id": "c2", "statement": "Every LLM call, tool invocation, wait, retry and branch is visible in the UI and API, and agents can be composed with other tasks, branching, schedules, human approval and cancellation.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The page names human approval as a composition element without documenting reviewer roles or approval records.", "capability": "step-level-visibility-and-human-approval" }, { "id": "c3", "statement": "Users and groups control access to an Orkes Conductor cluster, with direct roles, group memberships and inherited resource permissions spanning workflows, tasks, secrets, environment variables, tags, domains, integrations, prompts and gateway services.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The page does not document audit logging, tenant isolation or approval workflows.", "capability": "rbac-over-agent-resources" } ], "limitations": [ "No audit-log or tenant-isolation documentation was found in the pages reviewed.", "Human approval is a workflow composition element rather than a policy-driven approval product." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Google ADK", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Listed as an SDK authoring path; supported versions are not stated." } ], "conflicts": [], "editorialNotes": "One entry for Orkes Conductor including Conductor Agents; the upstream open-source Conductor OSS project was not entered separately, and productType hybrid reflects the OSS engine plus Orkes' commercial platform." }, { "name": "Palo Alto Networks AI Access Security", "id": "palo-alto-ai-access-security", "companyName": "Palo Alto Networks", "companyUrl": "https://www.paloaltonetworks.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "runtime-controls", "data-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Network security engineer", "SASE administrator" ], "problemsSolved": [ "GenAI app usage discovery", "policy control over AI app access", "prompt and upload data-leak prevention" ], "sources": [ { "id": "s1", "url": "https://docs.paloaltonetworks.com/ai-access-security/getting-started/introducing-ai-access-security", "title": "Introducing AI Access Security", "publisher": "Palo Alto Networks", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the product's purpose: identifying GenAI apps via an extensive app dictionary and mitigating prompt/file-upload data leakage with access-control policy." }, { "id": "s2", "url": "https://docs.paloaltonetworks.com/ai-access-security/getting-started/whats-supported-with-ai-access-security", "title": "What's Supported with AI Access Security?", "publisher": "Palo Alto Networks", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States visibility into hundreds of GenAI apps grouped by predefined use cases, the Insights dashboard for discovery, and Enterprise DLP as the detection engine." } ], "claims": [ { "id": "c1", "statement": "Provides visibility into hundreds of GenAI applications grouped into predefined use cases, with an Insights dashboard for discovering activity and risky behaviour by app, user and use case.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Discovery is bounded by the vendor-maintained GenAI app list and by traffic traversing Palo Alto enforcement points.", "capability": "genai-app-dictionary-discovery" }, { "id": "c2", "statement": "The product targets inadvertent data leakage in prompts and file uploads, combining GenAI app identification with fine-grained access control policy rules and Enterprise DLP detection.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Documentation does not establish coverage of autonomous agent traffic or MCP-based flows.", "capability": "prompt-and-upload-dlp" } ], "limitations": [ "No evidence of agent-level inventory (owner, tools, permissions)", "Requires traffic to traverse Palo Alto SASE enforcement" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept alongside Netskope because the evidenced scopes differ: Palo Alto's documentation establishes GenAI app usage discovery, while Netskope claims broader AI asset mapping." }, { "name": "Patronus AI", "id": "patronus-ai", "companyName": "Patronus AI", "companyUrl": "https://www.patronus.ai", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI engineering lead", "Model risk analyst" ], "problemsSolved": [ "hallucination detection", "agent failure diagnosis", "red-team probing of AI applications", "synthetic test dataset generation" ], "sources": [ { "id": "s1", "url": "https://docs.patronus.ai/", "title": "What is Patronus AI?", "publisher": "Patronus AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States that the platform evaluates, monitors and improves LLM systems via tracing, logging and alerts, uses in-house evaluators (Lynx, Glider), supports LLM-as-judge, human-in-the-loop annotation, dataset generation, prompt management, and red-teaming algorithms." }, { "id": "s2", "url": "https://www.patronus.ai/", "title": "Patronus AI | Simulating the World's Intelligence", "publisher": "Patronus AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Marketing page describing digital world models that simulate agent actions across deep research, multi-turn dialogue, long-horizon and memory task categories, with vendor-stated parity and model-lift figures." } ], "claims": [ { "id": "c1", "statement": "Documentation describes evaluating and monitoring LLM and agent interactions in production through tracing, logging and alerts, with in-house evaluators such as Lynx and Glider for hallucination and unsafe output, LLM-as-judge with custom criteria, human-in-the-loop annotations and dataset generation.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation overview does not state whether tool calls or multi-agent handoffs are captured as distinct spans.", "capability": "managed-evaluators-and-tracing" }, { "id": "c2", "statement": "Patronus documents red-teaming algorithms that automatically expose weaknesses in AI systems, alongside turnkey metrics covering RAG, agents, NLP and OWASP categories.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented attack taxonomy, coverage list or independent validation of the red-teaming results.", "capability": "red-teaming" }, { "id": "c3", "statement": "The company's product page states that digital world models predict and simulate agent actions in digital workflows, covering multi-turn dialogue, long-horizon tasks spanning days to months, memory and UI/UX navigation.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Simulation figures such as feature parity and model lift are vendor-stated on a marketing page with no methodology or third-party verification.", "capability": "agent-simulation" } ], "limitations": [ "No documented audit logging, retention configuration, RBAC or policy gating on either page fetched.", "Company positioning has shifted toward simulation/world models, so the evaluation product scope on the marketing site differs from the docs." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "description": "Docs present Patronus as an evaluation and monitoring platform for customer LLM systems, while the current marketing homepage foregrounds simulation environments and world models for agent training.", "sourceIds": [ "s1", "s2" ] } ], "editorialNotes": "Kept one entry covering the evaluation platform and simulation positioning; noted the docs/marketing divergence as a conflict rather than splitting into two products." }, { "name": "Permit MCP Gateway", "id": "permit-mcp-gateway", "companyName": "Permit.io", "companyUrl": "https://www.permit.io/", "primaryCategoryId": "runtime-controls", "secondaryCategoryIds": [ "identity-access" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "IAM manager", "Application security lead", "Platform engineer" ], "problemsSolved": [ "per-tool-call authorization", "binding agent actions to a human identity", "MCP audit trail" ], "sources": [ { "id": "s1", "url": "https://docs.permit.io/permit-mcp-gateway/overview/", "title": "Permit MCP Gateway Overview", "publisher": "Permit.io", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the gateway as an enforcement proxy that authenticates users, authorizes agent tool calls against fine-grained policy, collects consent and logs every decision." }, { "id": "s2", "url": "https://www.permit.io/mcp-gateway", "title": "Permit MCP Gateway | Drop-in Trust for AI Agents", "publisher": "Permit.io", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the gateway is a drop-in proxy for any MCP server, enforcing OPA-based RBAC/ABAC/ReBAC policy on every tool call with IdP-based authentication." } ], "claims": [ { "id": "c1", "statement": "Every MCP tool call proxied through the gateway is authorized in real time against fine-grained Permit.io policy, and denied calls are blocked before reaching the upstream server with a permission error.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation does not establish enforcement throughput, latency, or behaviour when the policy service is unreachable.", "capability": "per-tool-call-authorization" }, { "id": "c2", "statement": "The gateway authenticates users via SSO/OAuth, collects explicit consent, and logs each decision with agent identity, human identity, tool name, MCP server, timestamp and result.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The retention period and export format of decision logs are not documented on the fetched page.", "capability": "consent-and-audit" }, { "id": "c3", "statement": "Existing MCP clients are redirected to a hosted gateway URL that proxies to the upstream MCP server without modifying the server or the tools.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "The product page asserts compatibility with a long list of third-party MCP servers; those specific integrations were not independently verified.", "capability": "drop-in-proxy" } ], "limitations": [ "Fetched evidence covers a hosted gateway; self-hosted deployment was not documented on these pages" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "IL", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Distinct from Permit.io's general-purpose authorization service: this entry covers only the MCP enforcement proxy." }, { "name": "Pillar Security", "id": "pillar-security", "companyName": "Pillar Security", "companyUrl": "https://www.pillar.security/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "agent-discovery", "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AppSec lead", "AI security architect" ], "problemsSolved": [ "runtime detection of tool poisoning and injection", "tool-call schema and scope validation", "agent inventory and attack surface mapping" ], "sources": [ { "id": "s1", "url": "https://www.pillar.security/", "title": "Pillar Security | Securing the Agentic Workforce", "publisher": "Pillar Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes one platform to discover, test, secure and govern AI agents, with agentless inventory of agents, models, MCP servers and skills, attack surface mapping and adaptive runtime guardrails." }, { "id": "s2", "url": "https://www.pillar.security/solutions", "title": "AI Runtime Protection", "publisher": "Pillar Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States runtime protection monitors every prompt, tool call and command for prompt injection, tool poisoning and data exfiltration, validates tool call schemas and permission scope, and enforces policy on every production interaction." } ], "claims": [ { "id": "c1", "statement": "Pillar states its runtime protection monitors every prompt, tool call and command to detect prompt injection, tool poisoning and data exfiltration, and validates that tool calls match their declared schemas.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "No reference documentation, detection benchmarks or supported framework list was fetched.", "capability": "runtime-tool-call-monitoring" }, { "id": "c2", "statement": "The solutions page states Pillar validates that agents operate within their intended permission scope, flags deviations from declared intent, and intercepts poisoned instructions in inter-agent handoffs.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Enforcement mode (alert versus block) per capability is not specified.", "capability": "permission-scope-deviation" }, { "id": "c3", "statement": "Pillar describes cataloguing agents, models, prompts, frameworks, tools, MCP servers, skills and coding agents through agentless integrations, including shadow AI, and mapping how agents, tools, permissions and data sources connect.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Integration coverage and discovery method are not documented.", "capability": "lifecycle-inventory" } ], "limitations": [ "Evidence limited to vendor product pages", "No documented availability stage" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "IL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Pillar also publishes the SAIL framework, which is research content rather than a product and was not counted separately." }, { "name": "Planview Agent Resource Management", "id": "planview-agent-resource-management", "companyName": "Planview", "companyUrl": "https://www.planview.com", "primaryCategoryId": "work-delegation", "secondaryCategoryIds": [ "ai-risk" ], "productType": "commercial", "availability": "announced", "buyerRoles": [ "PMO leader", "Resource manager", "CIO" ], "problemsSolved": [ "blended human and agent capacity planning", "agent cost visibility", "human accountability for agent work" ], "sources": [ { "id": "s1", "url": "https://newsroom.planview.com/planview-launches-agent-resource-management-redefining-portfolio-resource-management-for-the-ai-era/", "title": "Planview Launches Agent Resource Management", "publisher": "Planview", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-05-04", "supports": "Announces one system to plan, predict, govern and optimize human and AI agent resources, with cost visibility and accountability to a human decision-maker." }, { "id": "s2", "url": "https://newsroom.planview.com/planview-closes-the-gap-between-strategic-intent-and-business-outcomes-in-the-age-of-ai/", "title": "Planview Closes the Gap Between Strategic Intent and Business Outcomes in the Age of AI", "publisher": "Planview", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2026-06-16", "supports": "States governance must be built for speed for humans and agents alike and introduces the Outcome Intelligence Graph linking decisions to resources and results." } ], "claims": [ { "id": "c1", "statement": "Planview announced agent resource management on 4 May 2026, extending its resource model so enterprises can plan, predict, govern and optimize both human and AI agent resources in one system and see who or what is doing the work and at what cost.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Press release; it does not state release timing, packaging, which Planview products carry the capability, or how agents report their work.", "capability": "blended-resource-planning" }, { "id": "c2", "statement": "The capability is described as holding every action accountable to a human decision-maker across a blended human and agent workforce.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documentation reviewed on how accountability is recorded, enforced or audited.", "capability": "human-accountability" }, { "id": "c3", "statement": "Planview's June 2026 announcement frames governance built for both humans and agents as a requirement of agentic-era portfolio management, alongside an Outcome Intelligence Graph connecting decisions to the resources behind them.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Strategy-level announcement; capability availability is not established.", "capability": "agentic-governance-direction" } ], "limitations": [ "No product documentation or GA confirmation found for agent resource management", "Mechanism for capturing agent effort and cost is unexplained" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Availability set to announced because only press releases were verifiable; product documentation pages for this capability were not reachable." }, { "name": "Presidio", "id": "presidio", "companyName": "Presidio community (originated at Microsoft)", "companyUrl": "https://github.com/microsoft/presidio", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Software engineer", "Privacy engineer", "Data scientist" ], "problemsSolved": [ "PII detection in unstructured text", "anonymization and redaction", "de-identification in images" ], "sources": [ { "id": "s1", "url": "https://github.com/microsoft/presidio", "title": "microsoft/presidio: An open-source framework for detecting PII", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Presidio as a data protection and de-identification SDK with identification and anonymization modules, predefined and custom recognizers, image redaction, and MIT license." }, { "id": "s2", "url": "https://data-privacy-stack.github.io/presidio/", "title": "Presidio", "publisher": "Presidio project", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Current documentation site listing PII identification in text and structured data, anonymization operators and image redaction, and stating the project is transitioning to community ownership." } ], "claims": [ { "id": "c1", "statement": "Provides identification and anonymization modules covering PII de-identification in text, images and structured or semi-structured data, using predefined or custom recognizers based on named entity recognition, regular expressions, rule-based logic and checksums.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Neither page states accuracy benchmarks or production support commitments.", "capability": "pii-detection-and-anonymization" }, { "id": "c2", "statement": "The repository is published under the MIT license.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Licensing of any downstream hosted service is not covered.", "capability": "mit-license" }, { "id": "c3", "statement": "The project documentation states Presidio is transitioning to a community-owned project and has moved to a new documentation site.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Governance structure, maintainers and long-term support commitments are not stated.", "capability": "community-ownership-transition" } ], "limitations": [ "No access control, lineage, consent or audit capability; it is a detection and anonymization library only.", "Ownership transition means the future maintenance model is uncertain.", "The old microsoft.github.io/presidio docs URL now redirects, which can confuse procurement." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "self_hosted" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Company fields reflect the community transition; kept a Microsoft attribution because the canonical repository is still microsoft/presidio. A fork at pvcy/presidio appeared in search results and was not treated as the canonical project." }, { "name": "Prisma AIRS AI Runtime Security", "id": "prisma-airs-ai-runtime-security", "companyName": "Palo Alto Networks", "companyUrl": "https://www.paloaltonetworks.com/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "runtime-controls", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "Network security architect", "SOC manager" ], "problemsSolved": [ "prompt injection and jailbreak blocking", "sensitive data leakage in AI traffic", "agent and MCP threat detection" ], "sources": [ { "id": "s1", "url": "https://docs.paloaltonetworks.com/ai-runtime-security", "title": "Prisma AIRS", "publisher": "Palo Alto Networks", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documentation hub describing AI Runtime API and firewall protection for AI apps, models, data and agents, listing MCP Threat Detection, Command Prompt Injection Detection, Tool Chaining Attack in agent scans and Privilege Misuse Detection for AI agents." }, { "id": "s2", "url": "https://www.paloaltonetworks.com/ai-security/ai-runtime-security", "title": "Prisma AIRS AI Runtime Security", "publisher": "Palo Alto Networks", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the service blocks prompt injections and data leaks, supports runtime security profiles and custom topics, and provides inspection of blocked prompts and violation sessions." } ], "claims": [ { "id": "c1", "statement": "Prisma AIRS provides an AI Runtime API to scan prompts and model responses programmatically and an AI Runtime firewall deployment that secures AI applications, models and datasets from threats such as prompt injections.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation index does not quantify detection accuracy or per-model coverage.", "capability": "api-and-network-scanning" }, { "id": "c2", "statement": "Documented release updates include MCP Threat Detection, Command Prompt Injection Detection, Tool Chaining Attack capability in agent scans, and Privilege Misuse Detection for AI agents.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Feature availability by region, tier and release date is not established from the index page alone.", "capability": "agent-specific-detections" }, { "id": "c3", "statement": "The product page describes creating and managing runtime security profiles, configuring custom topics, and inspecting blocked prompts and violation sessions by content type, threat category and severity.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "The product page does not state availability stage or licensing details.", "capability": "runtime-profiles-and-visibility" } ], "limitations": [ "Availability stage of individual agent detections not established", "Scope of coverage depends on deployment mode (API versus network)" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Prisma AIRS is a platform brand covering model scanning, posture and runtime; this entry is scoped to the AI Runtime Security component. Palo Alto's acquisition of Protect AI contributed model-scanning capabilities recorded elsewhere in the platform." }, { "name": "Private AI", "id": "private-ai", "companyName": "Private AI", "companyUrl": "https://private-ai.com", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Application security engineer", "Privacy engineer", "ML platform engineer" ], "problemsSolved": [ "PII removal before LLM calls", "re-identification after model response", "prompt de-identification for third-party models" ], "sources": [ { "id": "s1", "url": "https://docs.private-ai.com/reference/2.9/overview/", "title": "API Reference", "publisher": "Private AI", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the de-identification endpoint, enabled_classes, marker_format, unique_pii_markers, accuracy_mode, allow/block lists, fake entity generation and preserve_relationships options." }, { "id": "s2", "url": "https://private-ai.com/en/products/pgpt-api/", "title": "PrivateGPT Headless: The Privacy Layer for ChatGPT", "publisher": "Private AI", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2023-07-04", "supports": "Describes PrivateGPT Headless identifying and removing 50+ PII types before requests are sent to ChatGPT, with de-identification followed by re-identification." } ], "claims": [ { "id": "c1", "statement": "The API removes identifiers from one or more strings, with parameters controlling which PII classes are removed, the redaction marker format and whether markers uniquely identify each entity.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Reference does not state accuracy, throughput or supported languages.", "capability": "text-deidentification-api" }, { "id": "c2", "statement": "Supports generating fake entity replacements and preserving relationships so repeated mentions of the same entity map to the same synthetic value; both options are marked beta.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Beta status means behaviour may change; only 'standard' mode is supported.", "capability": "synthetic-entity-substitution" }, { "id": "c3", "statement": "PrivateGPT Headless is offered as a privacy layer that removes 50+ PII types before requests reach ChatGPT and re-identifies the data afterwards.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Product page dated 2023; current support for other model providers is not established.", "capability": "llm-privacy-proxy" } ], "limitations": [ "No access control, lineage, consent or audit-logging capability is documented.", "Publication date of the API reference is unknown, so version currency is unverified." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas" ], "geographies": [ "CA", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Distinct from the unrelated open-source 'PrivateGPT' project (docs.privategpt.dev) that appeared in search results; only Private AI's own docs were used." }, { "name": "Project Moonshot", "id": "aiverify-moonshot", "companyName": "AI Verify Foundation", "companyUrl": "https://aiverifyfoundation.sg", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "ai-risk" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI evaluation lead", "Compliance manager", "LLM application developer" ], "problemsSolved": [ "LLM safety benchmarking", "automated red-teaming", "test evidence for AI governance testing frameworks" ], "sources": [ { "id": "s1", "url": "https://github.com/aiverify-foundation/moonshot", "title": "aiverify-foundation/moonshot", "publisher": "AI Verify Foundation / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes benchmarking, red-teaming, IMDA starter kit workflows, attack modules and custom recipes." }, { "id": "s2", "url": "https://aiverifyfoundation.sg/project-moonshot/", "title": "Project Moonshot", "publisher": "AI Verify Foundation", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the toolkit, Python library plus web UI, evaluators, CI/CD integration and reporting." }, { "id": "s3", "url": "https://api.github.com/repos/aiverify-foundation/moonshot", "title": "GitHub REST API repository record", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Reports Apache-2.0 license, archived=false, last push 2026-06-10." } ], "claims": [ { "id": "c1", "statement": "The repository states Moonshot brings benchmarking and red-teaming together for LLM applications, tests bias, toxicity and hallucination alongside metrics such as accuracy and BLEU, supports guided workflows for IMDA's Starter Kit for LLM-based App Testing, and provides attack modules and custom recipes.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Benchmarks and attack modules are as good as their datasets; results are not certifications and dataset currency is not guaranteed.", "capability": "benchmark-and-redteam" }, { "id": "c2", "statement": "AI Verify Foundation states Project Moonshot is an open-source LLM evaluation toolkit with a Python library and web UI, pre-built evaluators and benchmark datasets, CI/CD pipeline integration and shareable reports.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Foundation page is promotional in tone and does not quantify coverage or validity of the evaluators.", "capability": "foundation-backing-and-ui" }, { "id": "c3", "statement": "GitHub metadata records Apache-2.0 licensing, a non-archived repository and last push June 2026.", "support": "documented", "sourceIds": [ "s3" ], "limitations": "Activity gap of several months; maintenance cadence unclear.", "capability": "license-and-activity" } ], "limitations": [ "Evaluation coverage depends on bundled datasets", "No agent-specific (multi-step tool use) evaluation evidence found" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "SG", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Related AI Verify testing framework not included separately; only the tool with a maintained repository is entered." }, { "name": "Promptfoo", "id": "promptfoo", "companyName": "Promptfoo", "companyUrl": "https://www.promptfoo.dev", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "agent-security" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "AI security engineer", "AI engineering lead" ], "problemsSolved": [ "adversarial testing of LLM applications", "prompt injection and jailbreak testing", "CI-based regression testing", "risk reporting for AI applications" ], "sources": [ { "id": "s1", "url": "https://www.promptfoo.dev/docs/red-team/", "title": "LLM red teaming guide (open source)", "publisher": "Promptfoo", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes open-source evaluation framework running inputs through an LLM and testing outputs, adversarial input generation, plugins, tool-based vulnerabilities, PII leaks from RAG context, deterministic and model-graded metrics, reporting and CI/CD usage." }, { "id": "s2", "url": "https://www.promptfoo.dev/docs/enterprise/", "title": "Promptfoo Enterprise - Secure LLM Application Testing", "publisher": "Promptfoo", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Promptfoo Enterprise is a fully managed SaaS edition and Enterprise On-Prem is self-hosted on AWS, Azure or GCP with a runner inside the network perimeter; lists RBAC, team management, network isolation and self-hosted sharing." } ], "claims": [ { "id": "c1", "statement": "Promptfoo documents generating a wide range of adversarial inputs via plugins covering prompt injection, jailbreaking, PII leakage from RAG context and tool-based vulnerabilities such as unauthorized data access and privilege escalation, then grading outputs with deterministic and model-graded metrics.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The guide does not quantify detection coverage or false-positive rates, and does not document capture of LLM spans or tool-call traces.", "capability": "adversarial-test-generation" }, { "id": "c2", "statement": "Promptfoo Enterprise is a hosted SaaS edition and Promptfoo Enterprise On-Prem is self-hosted on any major cloud with a dedicated runner inside the customer network; both list RBAC and team management.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Audit logs, SSO, retention settings and compliance certifications are not stated on the enterprise page.", "capability": "enterprise-rbac-and-onprem" }, { "id": "c3", "statement": "The red-teaming guide describes running scans continuously in CI/CD and feeding user-reported issues back into the red-team configuration on a review cadence.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented mechanism that blocks a release automatically; gating depends on the customer's pipeline.", "capability": "continuous-testing" } ], "limitations": [ "No documented audit logs, retention configuration, PII redaction or compliance certifications.", "Findings are test results rather than governance evidence records; no immutable report store is documented." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library", "saas", "self_hosted" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "productType hybrid: open-source CLI plus commercial SaaS and on-prem enterprise editions. Secondary agent-security assigned because the documented scope is adversarial security testing." }, { "name": "Protecto", "id": "protecto", "companyName": "Protecto", "companyUrl": "https://www.protecto.ai", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Application security engineer", "Privacy engineer", "AI platform lead" ], "problemsSolved": [ "tokenizing PII for LLM prompts", "policy-controlled re-identification", "reducing sensitive data in AI logs" ], "sources": [ { "id": "s1", "url": "https://docs.protecto.ai/core-concepts/tokenization-basics", "title": "Tokenization Basics - Protecto AI Documentation", "publisher": "Protecto", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-03-16", "supports": "Documents deterministic tokenization for LLM prompts and AI pipelines and policy-controlled, permission-gated unmasking that is never automatic." }, { "id": "s2", "url": "https://help.protecto.ai/getting-started/protecto-overview/introduction/protecto-vault/tokenization-apis", "title": "Tokenization APIs | Getting Started", "publisher": "Protecto", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2023-08-27", "supports": "Documents the masking API identifying and substituting PII with secure tokens and maintaining uniform tokens for the same value and token name." } ], "claims": [ { "id": "c1", "statement": "Replaces sensitive values with deterministic, non-sensitive tokens that can be stored, logged, shared and processed in AI and analytics workflows, including LLM prompts and AI pipelines.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Documentation does not state detection coverage or accuracy for automatically identifying sensitive values.", "capability": "deterministic-tokenization" }, { "id": "c2", "statement": "Tokens are reversible only when an explicit unmask request is made, the active policy allows unmasking and the caller has sufficient permissions; unmasking is never automatic.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The policy engine, role model and whether unmask events are logged are not documented on this page.", "capability": "policy-controlled-unmasking" }, { "id": "c3", "statement": "A masking API identifies and substitutes personally identifiable information in text with secure tokens, keeping tokens consistent for the same value and token name.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The help-centre page is dated 2023, so current API surface may differ.", "capability": "masking-api" } ], "limitations": [ "No documented data discovery across repositories, lineage, consent or audit logging.", "Vendor markets an 'AI data control plane for agentic AI' positioning that the fetched documentation does not substantiate." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "IN", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Homepage positions Protecto as an agentic-AI control plane; only tokenization and policy-gated unmasking were evidenced, so claims were scoped accordingly." }, { "name": "PwC Responsible AI Toolkit", "id": "pwc-responsible-ai-toolkit", "companyName": "PwC", "companyUrl": "https://www.pwc.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Chief Risk Officer", "Head of Compliance", "Chief Data Officer" ], "problemsSolved": [ "responsible AI governance design", "AI regulatory change monitoring", "bias assessment" ], "sources": [ { "id": "s1", "url": "https://www.pwc.com/sg/en/services/reimagine-digital/data-optimisation/what-is-responsible-ai.html", "title": "Responsible AI Toolkit", "publisher": "PwC Singapore", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the toolkit as customisable frameworks, tools and processes and offers a free Responsible AI Diagnostic." }, { "id": "s2", "url": "https://www.pwc.com/id/en/services/risk-assurance-services/data-analytics/responsible-ai.html", "title": "Responsible AI Toolkit", "publisher": "PwC Indonesia", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists governance, regulatory monitoring, risk detection and bias/fairness components of the toolkit." } ], "claims": [ { "id": "c1", "statement": "PwC describes the Responsible AI Toolkit as a suite of customisable frameworks, tools and processes covering strategy through execution, tailored to a client's AI maturity, and offers a free Responsible AI Diagnostic.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "The pages do not identify the specific tools, software, or artefacts included, nor who delivers them.", "capability": "toolkit-composition" }, { "id": "c2", "statement": "PwC states the toolkit covers enterprise-wide AI governance with roles and responsibilities across three lines of defence, traceability mechanisms, regulatory landscape monitoring, policy development, and bias/fairness assessment with corrective action.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Territory marketing page; no methodology document, control library, or client outcome evidence published.", "capability": "governance-and-compliance-components" } ], "limitations": [ "Contents of the 'toolkit' are not itemised publicly", "Delivery model and pricing undisclosed" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "SG", "ID", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Same named offering published by multiple PwC territory firms; treated as one entry." }, { "name": "PyRIT", "id": "pyrit", "companyName": "Microsoft", "companyUrl": "https://www.microsoft.com", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [ "agent-security" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI red team engineer", "Security engineer" ], "problemsSolved": [ "automated risk identification in generative AI systems", "repeatable adversarial testing" ], "sources": [ { "id": "s1", "url": "https://github.com/microsoft/PyRIT", "title": "microsoft/PyRIT: The Python Risk Identification Tool for generative AI", "publisher": "GitHub / Microsoft", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States PyRIT is an open source framework to proactively identify risks in generative AI systems, MIT licensed, public, with release v1.1.0 marked latest dated Sep 4 2026." }, { "id": "s2", "url": "https://github.com/Azure/PyRIT", "title": "Python Risk Identification Tool for generative AI (PyRIT)", "publisher": "GitHub / Microsoft", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Former Azure-org repository, archived by the owner on Mar 27 2026 and read-only, with a notice that PyRIT has moved to github.com/microsoft/PyRIT." } ], "claims": [ { "id": "c1", "statement": "The repository describes PyRIT as an open-source framework built to help security professionals and engineers proactively identify risks in generative AI systems, released under the MIT license.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The repository landing page does not enumerate attack techniques, scoring, datasets or captured artefacts.", "capability": "risk-identification-framework" }, { "id": "c2", "statement": "The maintained repository is microsoft/PyRIT, with release v1.1.0 listed as latest on 4 September 2026, while the former Azure/PyRIT repository was archived on 27 March 2026 and is read-only.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Release listing establishes activity but not feature scope or support commitments.", "capability": "maintenance-status" } ], "limitations": [ "Evidence for specific probe types, scorers and orchestrators was not captured from the pages fetched; only framework purpose and licensing are established.", "Purely a developer/red-team library: no governance features documented." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Repository relocated from the Azure org to the microsoft org; both fetched and the archive notice recorded. Feature-level claims deliberately limited to what the landing pages state. DEDUP FLAG: also present in services-oss.json. Recommend keeping one entry under evaluation (adversarial testing framework). Also surfaced in the open-source pass; evaluation kept as primary." }, { "name": "Ragas", "id": "ragas", "companyName": "VibrantLabs", "companyUrl": "https://github.com/vibrantlabsai", "primaryCategoryId": "evaluation", "secondaryCategoryIds": [], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "AI engineer", "ML practitioner" ], "problemsSolved": [ "RAG answer and retrieval quality measurement", "synthetic test set generation", "experiment-based comparison of application changes" ], "sources": [ { "id": "s1", "url": "https://docs.ragas.io/en/stable/", "title": "Ragas documentation", "publisher": "Ragas", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Ragas as a library for systematic evaluation loops, experiments-first approach, LLM-driven and custom metrics, dataset management, result tracking, and LangChain/LlamaIndex integrations; contact email founders@vibrantlabs.com." }, { "id": "s2", "url": "https://github.com/explodinggradients/ragas", "title": "vibrantlabsai/ragas: Supercharge Your LLM Application Evaluations", "publisher": "GitHub / VibrantLabs", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Repository now owned by vibrantlabsai; documents objective metrics, automatic test dataset creation, production-aligned test set generation, use of production data, custom metric example with score value and reason, and pip installation." } ], "claims": [ { "id": "c1", "statement": "Ragas documents an experiments-first workflow combining LLM-driven metrics, custom metrics defined with decorators, built-in dataset management and result tracking for LLM applications.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs do not state that spans, tool calls or multi-agent handoffs are captured; scope is dataset-level scoring.", "capability": "metric-and-experiment-workflow" }, { "id": "c2", "statement": "The repository states Ragas automatically creates test datasets covering a range of scenarios and supports production-aligned test set generation, with objective LLM-based and traditional metrics.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "No published validation of generated test-set representativeness.", "capability": "test-data-generation" }, { "id": "c3", "statement": "The canonical repository resolves to the vibrantlabsai organization and the library is installed from PyPI or source; no enterprise tier or hosted deployment is documented.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "License text and any commercial offering were not established from the pages fetched.", "capability": "ownership-and-distribution" } ], "limitations": [ "Developer library only: no audit records, retention, access control or policy gating.", "Vendor entity is thinly documented; company website not stated on the pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "LangChain", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Listed as a supported framework integration; depth not documented." } ], "conflicts": [ { "description": "The historical repository path explodinggradients/ragas now resolves to the vibrantlabsai organization, so ownership naming differs between the docs site and the repository.", "sourceIds": [ "s1", "s2" ] } ], "editorialNotes": "Recorded ownership move from the explodinggradients org to vibrantlabsai as observed on the fetched repository page." }, { "name": "Resaro Approved Intelligence", "id": "resaro-approved-intelligence", "companyName": "Resaro", "companyUrl": "https://resaro.ai", "primaryCategoryId": "services", "secondaryCategoryIds": [ "evaluation", "ai-risk" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Head of AI assurance", "Programme director", "Chief Risk Officer" ], "problemsSolved": [ "third-party AI evaluation", "deployment assurance evidence", "continuous post-deployment testing" ], "sources": [ { "id": "s1", "url": "https://resaro.ai/", "title": "Resaro - Approved Intelligence", "publisher": "Resaro", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the Approved Intelligence Platform, TEVV workflows run in the client environment, and continuous evaluation after deployment." }, { "id": "s2", "url": "https://resaro.ai/insights/articles/resaro-joins-global-ai-assurance-pilot-by-ai-verify-foundation-in-singapore", "title": "Resaro joins Global AI Assurance Pilot", "publisher": "Resaro", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States participation in the AI Verify Foundation and IMDA Global AI Assurance Pilot and the firm's Singapore/Germany footprint." } ], "claims": [ { "id": "c1", "statement": "Resaro states clients configure scenario-specific synthetic data and test packages and run TEVV workflows in their own environment rather than a remote lab, with continuous evaluation continuing after deployment.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product/service page; no published methodology, benchmark list, or example evidence pack.", "capability": "tevv-in-client-environment" }, { "id": "c2", "statement": "Resaro's article states it joined the Global AI Assurance Pilot launched by AI Verify Foundation and IMDA in Singapore, and describes itself as an AI assurance testing company with offices in Germany and Singapore.", "support": "independently_supported", "sourceIds": [ "s2" ], "limitations": "Company-authored article about a government-backed pilot; participation is not equivalent to accreditation or approval by IMDA.", "capability": "assurance-pilot-participation" } ], "limitations": [ "No independent accreditation evidenced", "Pilot participation verified only from the company's own article" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "SG", "DE", "EU" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Approved Intelligence is trademarked by Resaro; the platform is delivered as part of an assurance service, so productType is service rather than commercial software." }, { "name": "Saidot", "id": "saidot", "companyName": "Saidot", "companyUrl": "https://www.saidot.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI governance lead", "Compliance officer", "Product manager" ], "problemsSolved": [ "blank-sheet risk and control definition", "keeping governance current as AI changes", "AI inventory fragmentation" ], "sources": [ { "id": "s1", "url": "https://www.saidot.ai/product", "title": "AI governance platform: Inventory, risk, agents, and more", "publisher": "Saidot", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the knowledge-graph model linking systems, models, agents, datasets, risks, controls and policies, with automatic recommendation of risks, controls and policy requirements." }, { "id": "s2", "url": "https://help.saidot.ai/knowledge-base/getting-started-with-saidot", "title": "Getting started with Saidot", "publisher": "Saidot", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists governance workflows and tasks, policy management, risk management, evaluation and monitoring, reporting, lifecycle-based governance, Saidot Library, Saidot Governance, Saidot Graph, automated metadata generation and automated risk inheritance." } ], "claims": [ { "id": "c1", "statement": "Product page states governance is organised as a directed graph of systems, models, agents, datasets, risks, controls and policies, and that the platform recommends relevant risks, controls and policy requirements based on registered classifications and components.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page; recommendation logic, curation process and content coverage are not documented publicly.", "capability": "knowledge-graph-recommendations" }, { "id": "c2", "statement": "Help-centre documentation describes Saidot Library (curated knowledge base of policies, risks, models, products and evaluations), Saidot Governance (operationalising requirements per AI product), the Saidot Graph, automated metadata generation, automated risk identification and inheritance, evaluation and monitoring, reporting and a centralised AI inventory.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Knowledge-base overview names modules and capabilities but does not detail evaluation methods or supported connectors.", "capability": "governance-modules-and-library" } ], "limitations": [ "A public pricing page exists but was not fetched, so no pricing recorded.", "Evaluation/monitoring depth was not verifiable from the pages read." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Saidot is Finnish; HQ not stated on fetched pages so geography left empty." }, { "name": "SailPoint Agent Identity Security", "id": "sailpoint-agent-identity-security", "companyName": "SailPoint", "companyUrl": "https://www.sailpoint.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "ai-risk" ], "productType": "commercial", "availability": "unknown", "buyerRoles": [ "Identity governance lead", "IAM manager", "CISO" ], "problemsSolved": [ "registering AI agents as governed identities", "visibility into shadow AI tools", "governing service accounts, bots and RPAs" ], "sources": [ { "id": "s1", "url": "https://www.sailpoint.com/products/agent-identity-security", "title": "Agent Identity Security: Take control of AI agents", "publisher": "SailPoint", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-07-22", "supports": "Describes agent aggregation from cloud and agent platforms, automatic onboarding, unique identity registration and shadow AI visibility." }, { "id": "s2", "url": "https://www.sailpoint.com/products/machine-identity-security", "title": "Machine Identity Security - Products - SailPoint", "publisher": "SailPoint", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-08-04", "supports": "Describes discovery and governance of service accounts, bots, RPAs and APIs, including accounts created outside provisioning workflows." } ], "claims": [ { "id": "c1", "statement": "Agent Identity Security aggregates AI agents from AWS, Azure, GCP, Salesforce and Microsoft Copilot Studio, automatically onboards them and registers each with a unique identity plus business and access context.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; connector behaviour and coverage limits are not documented.", "capability": "agent-aggregation-and-registration" }, { "id": "c2", "statement": "SailPoint Machine Identity Security discovers and governs service accounts, bots, RPAs and other non-human identities, including accounts created outside formal provisioning workflows.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "The page does not describe credential, token or certificate issuance.", "capability": "machine-identity-governance" }, { "id": "c3", "statement": "The product provides visibility into hidden interactions involving unmonitored AI tools and supports remediation of shadow AI risk.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Detection method for unmonitored AI tools is not described.", "capability": "shadow-ai-visibility" } ], "limitations": [ "Availability stage and pricing not stated on fetched pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Two SailPoint modules (Agent Identity Security, Machine Identity Security) merged into one entry because they share the Identity Security Cloud buyer; machine identity claims cited separately." }, { "name": "Securiti Gencore AI", "id": "securiti-gencore-ai", "companyName": "Securiti", "companyUrl": "https://securiti.ai", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [ "runtime-controls", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Chief data officer", "CISO", "AI platform lead" ], "problemsSolved": [ "permission-aware RAG", "sensitive data redaction in AI pipelines", "AI data lineage" ], "sources": [ { "id": "s1", "url": "https://securiti.ai/gencore/", "title": "Gencore AI", "publisher": "Securiti", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2025-11-24", "supports": "Describes Gencore AI on the DataAI Command Platform: unstructured cataloging with sensitivity and entitlements, automatic detection and redaction in AI pipelines, permission-aware embeddings in protected vector DBs, and full GenAI pipeline lineage." }, { "id": "s2", "url": "https://securiti.ai/products/data-governance/", "title": "Data Governance Platform - Complete Solution - Securiti.ai", "publisher": "Securiti", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2025-07-22", "supports": "Lists the platform's governance areas including unstructured data governance, data access governance, AI governance, data catalog and data lineage." } ], "claims": [ { "id": "c1", "statement": "Creates and loads permission-aware custom embeddings into protected vector databases so AI copilots respect enterprise entitlements.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; does not document which source systems' entitlements are honoured or how permission changes propagate to existing embeddings.", "capability": "permission-aware-embeddings" }, { "id": "c2", "statement": "Automatically detects and redacts sensitive data within AI pipelines, with optional dynamic masking to comply with enterprise policy.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No documented entity coverage, accuracy or language support.", "capability": "in-pipeline-redaction" }, { "id": "c3", "statement": "Provides lineage and visibility across GenAI pipelines, including which files and embeddings reside in which vector databases.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Lineage depth and retention are not specified.", "capability": "genai-pipeline-lineage" }, { "id": "c4", "statement": "The wider Securiti platform lists data discovery and classification, unstructured data governance, data access governance, AI governance, data catalog and data lineage as product areas.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "The page names these areas without describing mechanisms, and does not state availability.", "capability": "data-governance-suite" } ], "limitations": [ "docs.securiti.ai did not load during review, so claims are product-page level rather than documented.", "An 'Agent Commander' offering with Veeam is promoted on the same pages but was not verifiable as generally available." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [ { "description": "Securiti pages promote 'Agent Commander' as an integrated Veeam + Securiti solution for scaling safe AI agents, but no product or documentation page establishing its availability was fetched.", "sourceIds": [ "s1", "s2" ] } ], "editorialNotes": "Entered as Gencore AI rather than the whole Securiti Data Command Center, since Gencore is the AI-specific product with permission-aware retrieval evidence." }, { "name": "ServiceNow AI Control Tower", "id": "servicenow-ai-control-tower", "companyName": "ServiceNow", "companyUrl": "https://www.servicenow.com/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "agent-discovery", "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI steward", "Risk and compliance manager", "IT service owner" ], "problemsSolved": [ "unknown AI asset estate", "no single owner for AI assets", "AI lifecycle and risk oversight" ], "sources": [ { "id": "s1", "url": "https://www.servicenow.com/docs/r/intelligent-experiences/ai-control-tower/ai-governance.html", "title": "AI Control Tower dashboard", "publisher": "ServiceNow", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-03-12", "supports": "Documents the AI Control Tower dashboard (AI status, inventory, metrics) and per-persona navigation for AI steward, product/asset owner, workspace user and risk and compliance user; Australia release, updated March 12, 2026." }, { "id": "s2", "url": "https://www.servicenow.com/products/ai-control-tower.html", "title": "AI Control Tower — ServiceNow", "publisher": "ServiceNow", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Claims automatic inventory of AI agents, models, copilots, MCP servers and datasets including shadow AI, CMDB-linked configuration items with lineage and ownership, AI identity/access and exposure tracking, prompt-injection blocking, risk assessment and compliance monitoring." } ], "claims": [ { "id": "c1", "statement": "ServiceNow product documentation for the Australia release describes an AI Control Tower dashboard showing AI status, AI inventory and AI metrics, with distinct navigation for AI steward, product/asset owner, workspace user and risk and compliance user personas.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs page describes dashboard and role visibility only; it does not detail discovery mechanisms or risk assessment content.", "capability": "persona-based-governance-dashboard" }, { "id": "c2", "statement": "Product page claims automatic inventory of first- and third-party AI agents, models, copilots, MCP servers and datasets, including shadow AI, with each asset enriched as a configuration item carrying lineage, ownership and relationships tied to the CMDB.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing page; discovery coverage, connectors and accuracy are not evidenced, and CMDB enrichment presumes ServiceNow platform adoption.", "capability": "ai-asset-discovery-and-cmdb" } ], "limitations": [ "Runtime security claims (prompt-injection blocking, least-privilege enforcement) come from marketing and were not verified in documentation.", "Requires the ServiceNow platform; standalone use is not evidenced." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "ServiceNow documents AI Control Tower per platform release (Yokohama, Australia); dated docs page used for versioned evidence." }, { "name": "Skyflow LLM Privacy Vault", "id": "skyflow-llm-privacy-vault", "companyName": "Skyflow", "companyUrl": "https://www.skyflow.com", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Privacy engineer", "Platform architect", "CISO" ], "problemsSolved": [ "PII isolation for LLM training and inference", "controlled re-identification", "sensitive data in RAG pipelines" ], "sources": [ { "id": "s1", "url": "https://docs.skyflow.com/docs/fundamentals/patterns/llm-privacy", "title": "Protect LLMs | Skyflow | Documentation", "publisher": "Skyflow", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-06-05", "supports": "Documents the LLM Privacy Vault pattern: de-identification at collection, Detect vault storing entities and tokens, controlled re-identification, and role-based access with permission-based partial disclosure." }, { "id": "s2", "url": "https://www.skyflow.com/product/skyflow-for-genai", "title": "GenAI Data Privacy and LLM Data Security", "publisher": "Skyflow", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-08-25", "supports": "States Skyflow for GenAI detects and redacts sensitive data across collection, training, fine-tuning, RAG and inference, and protects data flowing into and out of generative AI agents with fine-grained access controls." } ], "claims": [ { "id": "c1", "statement": "A Detect vault stores detected entities and their tokens so data can be de-identified at collection and re-identified under governance controls during training and inference.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation page describes the pattern, not detection accuracy or entity coverage.", "capability": "vault-deidentification" }, { "id": "c2", "statement": "Roles and access controls govern re-identification, with permissions allowing data to be fully or partially revealed only to validated users and authorized processes.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Does not document audit logging of re-identification events on this page.", "capability": "role-based-reidentification" }, { "id": "c3", "statement": "Skyflow for GenAI is presented as detecting and redacting sensitive data during collection, model training, fine-tuning, RAG and inference, including data flowing into and out of generative AI agents.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Product page; agent-specific integration mechanics are not described.", "capability": "genai-lifecycle-coverage" } ], "limitations": [ "No data discovery across existing repositories, lineage or consent management is documented.", "Adoption requires routing sensitive data through the vault, which the sources do not quantify in engineering effort." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Listed as the LLM Privacy Vault pattern/Skyflow for GenAI; the general-purpose PII vault was not counted as a separate AI governance product." }, { "name": "Snyk Agent Scan (formerly MCP-Scan)", "id": "snyk-agent-scan", "companyName": "Snyk", "companyUrl": "https://snyk.io", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "agent-discovery", "runtime-controls" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "CISO", "AppSec engineer", "Platform engineer" ], "problemsSolved": [ "MCP server risk scanning", "agent component inventory on endpoints", "prompt injection detection" ], "sources": [ { "id": "s1", "url": "https://github.com/snyk/agent-scan", "title": "snyk/agent-scan", "publisher": "Snyk / GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes agent component discovery and scanning, scan and background modes, the rename from MCP-Scan, and Apache-2.0 licensing." }, { "id": "s2", "url": "https://github.com/invariantlabs-ai/mcp-scan", "title": "invariantlabs-ai/mcp-scan (redirects to snyk/agent-scan)", "publisher": "GitHub", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Original MCP-Scan README describing static scan and proxy modes, tool poisoning, rug pull and cross-origin escalation detection." }, { "id": "s3", "url": "https://invariantlabs.ai/blog/introducing-mcp-scan", "title": "Introducing MCP-Scan", "publisher": "Invariant Labs", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes MCP-Scan's threat classes and tool pinning." }, { "id": "s4", "url": "https://snyk.io/news/snyk-acquires-invariant-labs-to-accelerate-agentic-ai-security-innovation/", "title": "Snyk acquires Invariant Labs", "publisher": "Snyk", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Confirms the acquisition and the MCP/tool-poisoning research it brings into Snyk Labs." } ], "claims": [ { "id": "c1", "statement": "The repository states the tool discovers installed agent components including harnesses, MCP servers and skills and scans MCP servers, tools, prompts, resources and skills for prompt injections, sensitive data handling and malware payloads hidden in natural language.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Detection quality is not quantified; local checks are supplemented by the Agent Scan API, so fully offline coverage is unclear.", "capability": "agent-component-scanning" }, { "id": "c2", "statement": "Earlier project documentation describes a proxy mode that monitors MCP connections in real time, checks tool calls, applies data-flow constraints and detects PII and indirect prompt injection, plus tool pinning that hashes tools to detect rug-pull changes.", "support": "documented", "sourceIds": [ "s2", "s3" ], "limitations": "Proxy-mode guardrailing relied on Invariant Guardrailing models via an API; feature naming changed after the rename to Agent Scan.", "capability": "runtime-proxy-and-pinning" }, { "id": "c3", "statement": "Snyk announced its acquisition of Invariant Labs, adding research on MCP vulnerabilities, tool poisoning and runtime detection to Snyk Labs and its AI Trust Platform.", "support": "independently_supported", "sourceIds": [ "s4" ], "limitations": "Vendor announcement; it does not state the licensing or long-term open-source commitment for the scanner.", "capability": "ownership-change" } ], "limitations": [ "Repository redirect means older documentation and the current tool name diverge", "Cloud API dependency for some checks" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "MCP-Scan", "mcp-scan" ], "parentCompany": "Snyk", "integrations": [], "conflicts": [], "editorialNotes": "GitHub API confirms invariantlabs-ai/mcp-scan now redirects to snyk/agent-scan (Apache-2.0, active September 2026). Recorded under the current name with formerNames preserved." }, { "name": "SPIRE (SPIFFE Runtime Environment)", "id": "spire", "companyName": "SPIFFE / Cloud Native Computing Foundation", "companyUrl": "https://spiffe.io", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls" ], "productType": "open_source", "availability": "generally_available", "buyerRoles": [ "Platform engineer", "Security architect", "SRE" ], "problemsSolved": [ "replacing static workload credentials", "cross-platform workload authentication", "automatic identity rotation" ], "sources": [ { "id": "s1", "url": "https://github.com/spiffe/spire", "title": "GitHub - spiffe/spire: The SPIFFE Runtime Environment", "publisher": "SPIFFE community", "type": "official_repository", "accessedAt": "2026-09-06", "publishedAt": "2026-08-21", "supports": "States SPIRE attests workloads, issues SPIFFE IDs/SVIDs, supports mTLS, JWT and Envoy SDS, and is a CNCF graduated project." }, { "id": "s2", "url": "https://spiffe.io/docs/latest/spiffe-about/overview/", "title": "SPIFFE Overview", "publisher": "SPIFFE", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes SPIFFE as open standards for bootstrapping and issuing short-lived identity documents to services across environments." } ], "claims": [ { "id": "c1", "statement": "SPIRE exposes the SPIFFE Workload API to attest running software systems and issue them SPIFFE IDs and SVIDs, enabling mTLS or signed JWT authentication between workloads.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Repository README does not address AI-agent-specific governance, inventory or approval workflows.", "capability": "workload-attestation-and-svid-issuance" }, { "id": "c2", "statement": "SPIRE is maintained by the SPIFFE community and is a graduated project of the Cloud Native Computing Foundation.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Graduation status does not imply enterprise support or SLAs.", "capability": "cncf-graduated-governance" }, { "id": "c3", "statement": "SPIFFE specifications define short-lived cryptographic identity documents (SVIDs) issued via a simple API across heterogeneous environments and organizational boundaries.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Specification-level description; implementation coverage varies by SPIFFE implementation.", "capability": "short-lived-identity-documents" } ], "limitations": [ "No license text was captured in this session; SPIRE is listed as open source on spiffe.io." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "library" ], "geographies": [ "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included as the reference open-source foundation that several commercial entries build on (Britive, Teleport, Defakto)." }, { "name": "SPLX AI Asset Management", "id": "splx-ai-asset-management", "companyName": "SPLX", "companyUrl": "https://splx.ai", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "agent-security", "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AI security lead", "AppSec manager" ], "problemsSolved": [ "AI asset inventory and AI-BOM", "MCP server discovery", "agentic workflow risk assessment" ], "sources": [ { "id": "s1", "url": "https://splx.ai/platform/ai-asset-management", "title": "AI Asset Management | Unified Inventory & Security", "publisher": "SPLX", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes LLM discovery, AI workflow scanning of code repos to reveal agents/tools/MCP servers, MCP server detection, and AI-BOM generation." }, { "id": "s2", "url": "https://splx.ai/blog/ai-asset-management-launch", "title": "SPLX Launches AI Asset Management", "publisher": "SPLX", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-09-10", "supports": "Launch announcement describing AI-BOM generation, vulnerability scanning and agentic workflow analysis, evolved from the open-source Agentic Radar tool." }, { "id": "s3", "url": "https://www.zscaler.com/press/zscaler-secures-enterprise-ai-lifecycle-acquisition-innovative-ai-security-pioneer-splx", "title": "Zscaler Acquires AI Security Pioneer SPLX", "publisher": "Zscaler", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-11-03", "supports": "Confirms Zscaler acquired SPLX to add shift-left AI asset discovery, automated red teaming and governance to its platform." }, { "id": "s4", "url": "https://splx.ai/", "title": "SPLX | End-to-End Security for AI", "publisher": "SPLX", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Homepage states SPLX is now part of Zscaler and still lists AI Asset Management as a platform module." } ], "claims": [ { "id": "c1", "statement": "Scans code repositories to map AI workflows and reveal agents, tools and MCP servers, and connects to cloud platforms, code repos and ML/AI platforms for a unified asset inventory.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; supported repository hosts, languages and agent frameworks are not enumerated.", "capability": "workflow-scanning" }, { "id": "c2", "statement": "AI Asset Management launched on 10 September 2025 combining AI-BOM generation, vulnerability scanning and agentic workflow analysis, building on SPLX's open-source Agentic Radar scanner.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Announcement does not quantify coverage or detection accuracy.", "capability": "launch-scope" }, { "id": "c3", "statement": "Zscaler announced the acquisition of SPLX on 3 November 2025, positioning shift-left AI asset discovery, red teaming and governance as extensions of the Zero Trust Exchange.", "support": "documented", "sourceIds": [ "s3", "s4" ], "limitations": "Neither source states how long SPLX remains purchasable as a standalone product or how the module will be repackaged.", "capability": "ownership-change" } ], "limitations": [ "Documentation site was unreachable during review", "Post-acquisition packaging and standalone availability unclear" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "EU", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [ "SplxAI" ], "parentCompany": "Zscaler", "integrations": [], "conflicts": [], "editorialNotes": "Formerly branded SplxAI. Recorded parentCompany Zscaler from the acquirer's own press release plus SPLX's homepage banner." }, { "name": "Straiker Defend AI", "id": "straiker-defend-ai", "companyName": "Straiker", "companyUrl": "https://www.straiker.ai/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "runtime-controls", "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "AI security lead", "SOC manager" ], "problemsSolved": [ "blocking prompt injection in agent workflows", "preventing destructive agent actions", "detecting data exfiltration by agents" ], "sources": [ { "id": "s1", "url": "https://www.straiker.ai/products/defend-ai", "title": "AI Agent Runtime Security and Protection | Straiker Defend AI", "publisher": "Straiker", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes runtime inspection of prompts, reasoning steps and tool calls, blocking of direct and indirect prompt injection, destructive actions and data exfiltration across coding, productivity and custom agents." }, { "id": "s2", "url": "https://www.straiker.ai/", "title": "Straiker | The Agentic AI Security Company", "publisher": "Straiker", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States the platform provides visibility and control across enterprise AI agents, shuts down rogue and risky agents and connections in real time, and that Ascend AI tests for prompt injection, goal hijacking, tool misuse and inter-agent manipulation." } ], "claims": [ { "id": "c1", "statement": "Straiker states Defend AI inspects every prompt, reasoning step and tool call in real time and blocks direct and indirect prompt injection, agent manipulation, jailbreaks and destructive actions such as file deletion and configuration changes.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No reference documentation, integration prerequisites or accuracy data were available.", "capability": "runtime-inspection-and-blocking" }, { "id": "c2", "statement": "Coverage is described across coding assistants such as Cursor, Claude Code and GitHub Copilot, productivity agents including Microsoft Copilot and ChatGPT Enterprise, and custom agents on AWS Bedrock and Azure AI Foundry.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Integration depth per platform is not documented.", "capability": "coverage-across-agent-types" }, { "id": "c3", "statement": "The company site states the platform provides visibility and control across enterprise AI agents, can shut down rogue and risky agents and connections in real time, and that a companion product tests for prompt injection, goal hijacking and tool misuse before deployment.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Shutdown mechanism and scope are not documented.", "capability": "rogue-agent-shutdown-and-testing" } ], "limitations": [ "Vendor marketing evidence only", "Availability stage not stated" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Two products (Defend AI runtime, Ascend AI testing) are marketed; this entry covers Defend AI, with testing noted as secondary evaluation capability rather than a separate listing." }, { "name": "Teleport Machine & Workload Identity", "id": "teleport-machine-workload-identity", "companyName": "Teleport", "companyUrl": "https://goteleport.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "runtime-controls", "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "Platform engineer", "Infrastructure security lead", "DevSecOps lead" ], "problemsSolved": [ "static credentials in automation and agents", "restricting agent actions on infrastructure", "auditing agent activity on infrastructure" ], "sources": [ { "id": "s1", "url": "https://goteleport.com/docs/machine-workload-identity/use-cases/ai-agents-mwi/", "title": "AI Agents with Machine & Workload Identity", "publisher": "Teleport", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents issuing agents their own identity, RBAC restriction to authorized actions and full logging of agent commands." }, { "id": "s2", "url": "https://goteleport.com/platform/machine-and-workload-identity/", "title": "Machine & Workload Identity for Agents, MCP, SSH", "publisher": "Teleport", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States issuance of short-lived certificates, tokens and SPIFFE-compatible identities to machines, workloads and AI agents via tbot." } ], "claims": [ { "id": "c1", "statement": "Teleport issues each AI agent its own identity and enforces access and privileges for agents accessing SSH servers, Kubernetes clusters, databases and MCP servers.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The docs page does not state GA status for the AI agent use case and invites design partnerships.", "capability": "agent-identity-issuance" }, { "id": "c2", "statement": "The platform issues and renews short-lived X.509 certificates, JWTs and SPIFFE-compatible workload identities for bots, CI jobs, containers, VMs, services and AI agents.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Platform page is marketing; per-protocol rotation intervals are not stated.", "capability": "short-lived-workload-credentials" }, { "id": "c3", "statement": "All queries, commands and requests executed by an agent are logged, providing visibility and auditability.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Retention, export and tamper-evidence properties are not described on the fetched page.", "capability": "agent-action-auditing" } ], "limitations": [ "No evidence of agent discovery or entitlement review campaigns." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Teleport has open-source components; productType recorded as hybrid because the fetched pages describe a commercial platform." }, { "name": "Temporal", "id": "temporal", "companyName": "Temporal Technologies", "companyUrl": "https://temporal.io", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "observability" ], "productType": "hybrid", "availability": "generally_available", "buyerRoles": [ "Platform engineering lead", "Staff backend engineer", "AI infrastructure architect" ], "problemsSolved": [ "reliable long-running agent execution", "human approval steps in automated workflows", "recovery from failures without losing state" ], "sources": [ { "id": "s1", "url": "https://docs.temporal.io/ai", "title": "Durable AI | Temporal Documentation", "publisher": "Temporal", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents durable stateful agent loops, automatic resumption after failure or multi-day human waits, human-in-the-loop and approval patterns using Signals, tool calling and MCP cookbook entries, and task queue priority/fairness for tenant workload isolation." }, { "id": "s2", "url": "https://temporal.io/product", "title": "Durable Execution Platform", "publisher": "Temporal", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Temporal as a durable execution platform with Temporal Cloud as a managed serverless offering, incorporation of human interactions in workflows, and the ability to inspect, replay and rewind every workflow execution." } ], "claims": [ { "id": "c1", "statement": "Supports long-running stateful agent loops that call LLMs and tools, wait on humans, and resume automatically after a crash, network timeout or multi-day wait.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation covers execution reliability, not model or data governance.", "capability": "durable-agent-loops" }, { "id": "c2", "statement": "Approval workflows block until external approval decisions are made, using Signals to capture approval data with metadata, and external sources including human actions can interact with workflows.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "There is no described approval UI, reviewer role model or policy engine; approvals are implemented by developers.", "capability": "human-approval-primitives" }, { "id": "c3", "statement": "Every workflow execution can be inspected, replayed and rewound step by step.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The page does not describe this as a tamper-evident audit log or state retention periods.", "capability": "execution-history-inspection" }, { "id": "c4", "statement": "Task queue priority and fairness can prioritise urgent work and keep one tenant from starving others, and worker configuration is used to route and isolate agent workloads.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "This is scheduling fairness, not a security tenant-isolation boundary.", "capability": "workload-isolation" } ], "limitations": [ "Governance is developer-only: no approval console, RBAC documentation for agent actions, or AI-specific audit trail in the pages reviewed.", "Not an AI product per se; AI capability is a documented usage pattern on a general-purpose orchestration engine." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "self_hosted", "saas", "library" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "OpenAI Agents SDK", "type": "documented_api", "sourceIds": [ "s1" ], "limitations": "Referenced through cookbook recipes; support level and version compatibility are not stated." } ], "conflicts": [], "editorialNotes": "Included because Temporal publishes a dedicated AI/agent documentation section, distinguishing it from generic workflow schedulers that were excluded." }, { "name": "Token Security NHI & AI Agent Security Platform", "id": "token-security", "companyName": "Token Security", "companyUrl": "https://www.token.security", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery" ], "productType": "commercial", "availability": "unknown", "buyerRoles": [ "CISO", "IAM manager", "Identity security lead" ], "problemsSolved": [ "NHI and agent inventory", "ownership and accountability for machine identities", "decommissioning orphaned agent identities" ], "sources": [ { "id": "s1", "url": "https://www.token.security/", "title": "Token Security: AI Agent & Non-Human Identity Security Platform", "publisher": "Token Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States continuous discovery of agents and NHIs, unified identity graph, ownership enforcement, lifecycle governance and automatic decommissioning." }, { "id": "s2", "url": "https://www.token.security/lp/ai-agent-identity-lifecycle-management-and-governance", "title": "AI Agent Identity Lifecycle Management - Token Security", "publisher": "Token Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes governing AI agents from creation to decommissioning, inventory of NHIs, ownership establishment, risk analysis and remediation." } ], "claims": [ { "id": "c1", "statement": "The platform continuously discovers every AI agent and NHI across on-prem, hybrid and cloud environments and correlates agents, humans, secrets, permissions and data in a unified identity graph.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Home page does not document connector coverage or discovery methods.", "capability": "agent-nhi-discovery" }, { "id": "c2", "statement": "Token manages the full AI agent identity lifecycle from creation through retirement, enforces ownership and automatically decommissions orphaned identities.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "The second source is a gated eBook landing page describing a framework rather than product reference documentation.", "capability": "agent-identity-lifecycle" } ], "limitations": [ "No official technical documentation fetched; availability stage unknown." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Company name 'Token Security' is unrelated to OAuth token vaulting products; no acquisition found." }, { "name": "Tonic Textual", "id": "tonic-textual", "companyName": "Tonic.ai", "companyUrl": "https://www.tonic.ai", "primaryCategoryId": "data-governance", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Data engineer", "Privacy engineer", "ML platform lead" ], "problemsSolved": [ "unstructured data de-identification", "safe training and RAG datasets", "prompt redaction" ], "sources": [ { "id": "s1", "url": "https://docs.tonic.ai/textual", "title": "Tonic Textual guide", "publisher": "Tonic.ai", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents dataset scanning for sensitive values, redaction or replacement, guided redaction with review, LLM-based review of detection quality, REST API for redaction and user management, and an MCP server for external AI agents." }, { "id": "s2", "url": "https://docs.tonic.ai/textual/tonic-textual-api/datasets-redaction", "title": "Datasets and redaction | Tonic Textual", "publisher": "Tonic.ai", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2024-12-18", "supports": "Documents creating and managing datasets, redacting strings and files via the API, and recording and reviewing redaction requests in the application." }, { "id": "s3", "url": "https://www.tonic.ai/products/textual", "title": "Tonic Textual | Unstructured Data De-Identification", "publisher": "Tonic.ai", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Textual redacts or synthesizes sensitive entities in text, audio, LLM prompts and RAG data using proprietary NER models, supports custom entity types, and lists RBAC and SSO as enterprise features." } ], "claims": [ { "id": "c1", "statement": "Textual scans dataset files to identify sensitive values and lets users redact or replace them, including guided redaction projects with human review before output files are downloaded.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "Documentation does not state detection accuracy or language coverage.", "capability": "detect-and-redact-unstructured" }, { "id": "c2", "statement": "A REST API and Python SDK redact text strings and manage datasets and user access, and a Textual MCP server can receive requests from an external AI agent.", "support": "documented", "sourceIds": [ "s1", "s2" ], "limitations": "The MCP server's authentication and scoping model is not described on these pages.", "capability": "sdk-and-mcp-access" }, { "id": "c3", "statement": "Role-based access controls and single sign-on are listed as enterprise security features of Textual.", "support": "vendor_claim", "sourceIds": [ "s3" ], "limitations": "Product page listing; no permission model documentation was fetched.", "capability": "rbac-and-sso" } ], "limitations": [ "No lineage, consent management or documented audit-log capability.", "Access control claims rest on a marketing page rather than documentation." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Kept separate from Tonic Structural (structured-data synthesis), which was out of scope for AI-specific governance." }, { "name": "trail", "id": "trail", "companyName": "trail", "companyUrl": "https://www.trail-ml.com/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI governance lead", "Compliance manager", "Data science lead" ], "problemsSolved": [ "scattered AI asset records", "repeated assessment effort", "EU AI Act role and risk classification" ], "sources": [ { "id": "s1", "url": "https://www.trail-ml.com/ai-governance", "title": "AI Governance Made Simple — trail", "publisher": "trail", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes centralized AI inventory and classification, EU AI Act risk/role classification, vendor assessment templates, curated frameworks and controls with automated mapping and evidence tracking." }, { "id": "s2", "url": "https://www.trail-ml.com/blog/the-ai-registry-with-superpowers", "title": "The AI Registry with Superpowers", "publisher": "trail", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the trail AI Registry: system of record for models, systems, agents, use cases and vendors, dependency mapping, ownership and governance status, and reuse of assessments and evidence." } ], "claims": [ { "id": "c1", "statement": "Product page states guided risk and role classification for AI systems under the EU AI Act plus prebuilt framework templates such as EU AI Act and ISO 42001, with automated mapping and evidence tracking for controls.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page. Guided classification is a vendor-designed questionnaire; it does not constitute a legal determination of EU AI Act obligations.", "capability": "eu-ai-act-classification-and-templates" }, { "id": "c2", "statement": "Vendor article describes a centralized registry of models, systems, agents, use cases and third-party vendors that maps and visualises dependencies between assets, tracks ownership and governance status, and allows reuse of prior assessments and evidence.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Vendor blog post updated 25.06.2026; automated registration from AI platforms is asserted without documented connector list.", "capability": "living-ai-registry-with-dependencies" }, { "id": "c3", "statement": "Product page states vendor and third-party AI solutions can be tracked and assessed using maintained templates for popular AI services.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Template inventory and update cadence are not evidenced.", "capability": "third-party-ai-assessment" } ], "limitations": [ "No public technical documentation was found; claims rest on vendor pages.", "Runtime enforcement material exists on the vendor blog but was not verified and is excluded here." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Company brands itself lowercase 'trail'; domain is trail-ml.com. Held back the separate 'AI agent governance runtime enforcement' blog claim as unverified." }, { "name": "Trustible AI Governance Platform", "id": "trustible", "companyName": "Trustible", "companyUrl": "https://trustible.ai/", "primaryCategoryId": "ai-risk", "secondaryCategoryIds": [ "agent-discovery" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "AI governance lead", "Compliance officer", "Legal counsel" ], "problemsSolved": [ "unstructured AI intake", "inconsistent risk triage", "regulatory change tracking" ], "sources": [ { "id": "s1", "url": "https://trustible.ai/platform-overview/", "title": "AI Governance Platform — Trustible", "publisher": "Trustible", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Sets out intake, manage risk, monitor and comply stages, unified AI inventory, curated risk taxonomies and framework mappings, control and audit evidence." }, { "id": "s2", "url": "https://trustible.ai/", "title": "Trustible — The AI Governance Platform for Regulated Enterprises", "publisher": "Trustible", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States single source of truth for AI use cases, models, agents and vendors; structured intake with automatic risk-based routing; inventory dashboard with risk distribution and framework readiness." } ], "claims": [ { "id": "c1", "statement": "Vendor platform page describes four stages: intake (use-case, model and vendor capture with risk-based triage), risk management (risk and impact assessments, mitigation and incident tracking), monitoring (AI metrics, regulatory and vendor alerts) and compliance (global framework mappings, control and audit evidence).", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product overview page; no technical documentation was found for assessment templates, metric collection method or evidence formats.", "capability": "lifecycle-intake-to-compliance" }, { "id": "c2", "statement": "Homepage states a single inventory covering every AI use case, model, agent and vendor, with owner and governance-role assignment and an inventory dashboard showing risk-level distribution, mitigation coverage and framework readiness.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Screenshots and counts shown are illustrative; the page does not evidence automated discovery of AI assets.", "capability": "unified-ai-inventory" } ], "limitations": [ "No public product documentation or API reference was located, so all capability claims rest on vendor pages.", "Deployment model and pricing not stated on fetched pages." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Positioned explicitly against repurposed GRC tools; no acquisitions or renames found." }, { "name": "TÜV SÜD ISO/IEC 42001 certification", "id": "tuv-sud-iso-42001-certification", "companyName": "TÜV SÜD", "companyUrl": "https://www.tuvsud.com", "primaryCategoryId": "services", "secondaryCategoryIds": [ "ai-risk" ], "productType": "service", "availability": "generally_available", "buyerRoles": [ "Quality manager", "Head of Compliance", "CISO" ], "problemsSolved": [ "ISO/IEC 42001 certification", "integrated management system audits", "AI risk assessment review" ], "sources": [ { "id": "s1", "url": "https://www.tuvsud.com/en/services/management-system-certification/iso-iec-42001", "title": "ISO IEC 42001 certification", "publisher": "TÜV SÜD", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States TÜV SÜD tests and certifies AIMS to ISO/IEC 42001 and offers integrated management system certification; references certifications 'under various accreditations' without naming one." }, { "id": "s2", "url": "https://www.tuvsud.com/en-us/services/auditing-and-system-certification/iso-iec-42001-artificial-intelligence-management-system", "title": "ISO/IEC 42001 Artificial Intelligence Management System", "publisher": "TÜV SÜD", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes the AIMS framework, transparency/fairness/risk focus and TÜV SÜD's certification service." } ], "claims": [ { "id": "c1", "statement": "TÜV SÜD states it tests and certifies AI management systems according to the requirements of ISO/IEC 42001 and can perform integrated management system certification evaluating multiple standards in a single audit.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No accreditation body or accreditation number is named on the page for the ISO/IEC 42001 scheme.", "capability": "certification-service" }, { "id": "c2", "statement": "TÜV SÜD's US page describes ISO/IEC 42001 as the first international AIMS standard and positions its service around responsible development, transparency, fairness and risk mitigation.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Descriptive standard summary; does not evidence audit depth, auditor competence criteria, or certificate scope wording.", "capability": "assessment-focus" } ], "limitations": [ "No named accreditation for the ISO/IEC 42001 scheme on the reviewed pages", "No EU AI Act notified-body claim evidenced" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [], "geographies": [ "DE", "EU", "US", "IN", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "TÜV SÜD's separate AI Quality Certification Program (AIQCP) training was held as a distinct offering to avoid double-counting the same vendor." }, { "name": "UiPath Maestro", "id": "uipath-maestro", "companyName": "UiPath", "companyUrl": "https://www.uipath.com", "primaryCategoryId": "agent-building", "secondaryCategoryIds": [ "work-delegation", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Automation CoE lead", "Process owner", "Enterprise architect" ], "problemsSolved": [ "orchestrating agents robots and people", "governed process change management", "long-running case work" ], "sources": [ { "id": "s1", "url": "https://docs.uipath.com/maestro/automation-cloud/latest/user-guide/overview", "title": "Maestro - Overview", "publisher": "UiPath", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": "2026-08-28", "supports": "Documents UiPath Maestro as a cloud-native orchestration platform unifying automation, AI agents and human interactions, with BPMN, Flow and Case Management modelling on one engine, DMN business rules, and combined process and automation data." }, { "id": "s2", "url": "https://www.uipath.com/product/maestro", "title": "UiPath Maestro: Business Orchestration with BPMN and AI Agents", "publisher": "UiPath", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Maestro coordinates AI agents, UiPath robots and people, integrates third-party agents, hands exceptions to people, and that improvements and decision tables are versioned and audited with role-based access and rollback." } ], "claims": [ { "id": "c1", "statement": "Maestro is a cloud-native orchestration platform that unifies automation, AI agents and human interactions, coordinating RPA bots, AI tools and people within a single process, using BPMN, Flow or Case Management models on a shared engine with DMN business rules.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "The overview documents no agent-building interface, SDK or tool library; agents are orchestrated rather than authored here.", "capability": "multi-actor-orchestration" }, { "id": "c2", "statement": "Improvements and decision tables are versioned and audited alongside the BPMN model, with role-based access, versioning, audit and rollback named as platform characteristics.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Stated on a product page; the audit record's contents, retention and export are not documented, and log streaming is not mentioned.", "capability": "versioning-and-audit-of-process-change" }, { "id": "c3", "statement": "Exceptions are handed to people, with AI agents reasoning, robots executing system work and people handling judgement calls; third-party agents, robots and people can be integrated.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Explicit human-in-the-loop approval gating of an agent's individual tool calls is not documented.", "capability": "human-exception-handling" }, { "id": "c4", "statement": "Per-agent permission scoping and tenant isolation for agent execution are not established by the pages reviewed.", "support": "not_found", "sourceIds": [ "s1", "s2" ], "limitations": "Absence of evidence; a Maestro agent-governance documentation page timed out during review.", "capability": "agent-level-permission-scoping" } ], "limitations": [ "Agent authoring happens in other UiPath products; Maestro's documented role is orchestration and governance of processes.", "Audit and RBAC claims come from a marketing page; the agent-governance documentation page did not load during review." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [ "US", "RO", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Entered Maestro specifically rather than 'UiPath agentic automation' generally, per the instruction to name the specific product for big-vendor platforms. Also surfaced in work-delegation research; agent-building kept as primary for the same reason." }, { "name": "Valence SaaS and AI Discovery", "id": "valence-saas-ai-discovery", "companyName": "Valence Security", "companyUrl": "https://www.valencesecurity.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "identity-access", "agent-security" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "SaaS security engineer", "IAM manager" ], "problemsSolved": [ "shadow SaaS and AI app inventory", "risky OAuth integration discovery", "AI agent posture in SaaS" ], "sources": [ { "id": "s1", "url": "https://www.valencesecurity.com/platform/saas-and-ai-discovery", "title": "SaaS Discovery - Secure All Your SaaS Apps", "publisher": "Valence Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes complete SaaS/AI app inventory covering shadow IT and AI tools, and continuous identification of OAuth tokens, API keys, connected apps and service accounts." }, { "id": "s2", "url": "https://www.valencesecurity.com/platform/ai-security-posture-management", "title": "AI Security Posture Management (AI-SPM)", "publisher": "Valence Security", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States Valence discovers AI-enabled SaaS tools and integrations, governs AI capabilities, and secures AI agents operating inside business applications." } ], "claims": [ { "id": "c1", "statement": "Continuously identifies OAuth tokens, API keys, connected apps and service accounts used to connect business-critical SaaS applications to third-party SaaS and AI applications.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Product page; no documentation reviewed on connector coverage, refresh interval, or detection method per SaaS vendor.", "capability": "oauth-and-nonhuman-discovery" }, { "id": "c2", "statement": "Discovers AI-enabled SaaS tools and integrations, and covers AI agents operating inside business applications as part of AI security posture management.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Does not name the specific agent platforms enumerated, nor distinguish detected-agent metadata from configuration posture checks.", "capability": "ai-feature-and-agent-posture" }, { "id": "c3", "statement": "Builds an inventory covering both sanctioned and unsanctioned SaaS apps, including non-SSO and local accounts.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Discovery scope for apps with no SaaS-side integration footprint is unstated.", "capability": "shadow-app-inventory" } ], "limitations": [ "Discovery is limited to SaaS tenants Valence is connected to", "No official docs reviewed listing supported AI agent platforms" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "IL", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Israeli-founded SSPM vendor. Kept as one entry combining the SaaS/AI discovery and AI-SPM pages, which describe the same discovery engine." }, { "name": "ValidMind", "id": "validmind", "companyName": "ValidMind", "companyUrl": "https://validmind.com/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "ai-risk", "evaluation" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Model risk manager", "Model validator", "Head of model governance" ], "problemsSolved": [ "manual model documentation", "slow independent validation cycles", "inventory and lifecycle oversight for AI models" ], "sources": [ { "id": "s1", "url": "https://docs.validmind.ai/about/overview-model-documentation.html", "title": "Automated testing & documentation — ValidMind", "publisher": "ValidMind", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the ValidMind Library: automated documentation and testing, upload to the ValidMind Platform for review and approval, test suites for statistical/AI/ML/LLM/agentic records, data quality, robustness and explainability assessment, template-based documentation generation." }, { "id": "s2", "url": "https://validmind.com/platform/", "title": "ValidMind AI Governance Platform | Model Risk Management", "publisher": "ValidMind", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "States model inventory and lifecycle tracking, version controls, standardized validation workflows with approvals and reporting, platform-agnostic support for statistical, AI and GenAI models." } ], "claims": [ { "id": "c1", "statement": "ValidMind documentation states the ValidMind Library automates model documentation and testing, uploads qualitative and quantitative test data to the ValidMind Platform for review and approval, runs test suites across statistical, AI/ML and LLM models and agentic systems, and assesses data quality, outcomes, robustness and explainability.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs describe the library's function and handoff; they do not evidence a model inventory or registry, nor validate that outputs meet a specific regulator's expectations.", "capability": "automated-testing-and-documentation" }, { "id": "c2", "statement": "Platform page states model inventory and lifecycle tracking, version controls, standardized validation workflows with approvals and reporting, and support for traditional statistical, AI, GenAI and bring-your-own-model deployments.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Marketing page; workflow configurability, RBAC detail and inventory scope are not documented there.", "capability": "inventory-workflows-approvals" } ], "limitations": [ "Financial-services customer claims seen in third-party coverage were not verified and are excluded.", "Self-hosted or private-cloud deployment not evidenced on pages fetched." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Two components (ValidMind Library, ValidMind Platform) treated as one product entry." }, { "name": "Velatir", "id": "velatir", "companyName": "Velatir", "companyUrl": "https://www.velatir.com", "primaryCategoryId": "work-delegation", "secondaryCategoryIds": [ "runtime-controls", "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Head of AI governance", "Compliance officer", "Platform engineering lead" ], "problemsSolved": [ "human approval gates for agent actions", "delegation boundaries for autonomous operations", "auditable AI decision records" ], "sources": [ { "id": "s1", "url": "https://docs.velatir.com/review-tasks/understanding-review-tasks", "title": "Understanding Review Tasks", "publisher": "Velatir", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents the review task lifecycle: creation, policy assessment, flow routing, human review, resolution, with risk levels and auto-approval for low risk." }, { "id": "s2", "url": "https://docs.velatir.com/integrations/mcp-server", "title": "MCP Server Integration", "publisher": "Velatir", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents an MCP server acting as a human-in-the-loop approval gate before agents execute sensitive operations, with request tracking and denial feedback." }, { "id": "s3", "url": "https://www.velatir.com/", "title": "Velatir - See how AI is being used, control what matters", "publisher": "Velatir", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Positions the product as a control layer giving accountable owners a frame to set rules while showing the organisation's AI estate." }, { "id": "s4", "url": "https://cyberinsider.co.uk/velatir-raises-e5m-for-ai-controls/", "title": "Velatir raises EUR 5m for AI controls", "publisher": "Cyber Insider", "type": "independent_reporting", "accessedAt": "2026-09-06", "publishedAt": "2026-08-21", "supports": "Reports Velatir is a Danish startup raising EUR 5m seed to expand a European control layer monitoring how employees and autonomous agents use AI." } ], "claims": [ { "id": "c1", "statement": "Each AI request needing oversight becomes a review task submitted via API, SDK or integration, is assessed against active policies, routed to an approval channel, decided by designated reviewers, and resolved with an audit trail; low-risk requests can auto-approve while medium/high-risk ones require human intervention.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Documentation describes the workflow model; it does not establish enforcement guarantees if an agent bypasses the SDK or MCP gate.", "capability": "review-task-lifecycle" }, { "id": "c2", "statement": "A Velatir MCP server provides an approval gate requiring human authorisation before MCP-compatible AI agents execute sensitive operations, with request tracking by ID and denial reasons fed back to the agent.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Applies only to tools routed through the Velatir MCP server or SDK.", "capability": "mcp-approval-gate" }, { "id": "c3", "statement": "Velatir is a Danish AI governance startup that raised a EUR 5 million seed round reported in August 2026 to expand a European control layer for monitoring enterprise AI use.", "support": "independently_supported", "sourceIds": [ "s4" ], "limitations": "Customer counts in the article are company-provided figures; funding does not evidence product capability.", "capability": "company-status" } ], "limitations": [ "Approval enforcement requires agents to call Velatir's SDK/MCP gate", "Marketing site also claims broad AI usage visibility that the reviewed docs do not detail" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "EU", "DK" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [ { "name": "Model Context Protocol (MCP)", "type": "documented_api", "sourceIds": [ "s2" ], "limitations": "Documented as a Velatir-provided MCP server; compatibility per client tool not enumerated." } ], "conflicts": [], "editorialNotes": "Placed in work-delegation because the documented core is routing agent work to accountable human reviewers; runtime-controls is secondary." }, { "name": "Verint Agent Factory", "id": "verint-agent-factory", "companyName": "Verint", "companyUrl": "https://www.verint.com", "primaryCategoryId": "work-delegation", "secondaryCategoryIds": [ "agent-building", "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Contact centre operations leader", "CX transformation lead", "Head of automation" ], "problemsSolved": [ "human/AI work routing in CX", "hybrid workforce orchestration", "prompt and model governance" ], "sources": [ { "id": "s1", "url": "https://www.verint.com/agent-factory/", "title": "Verint Agent Factory | AI Agent Orchestration", "publisher": "Verint", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes a single environment to build, govern and scale a hybrid workforce of human and AI agents, routing to humans when judgment is needed." }, { "id": "s2", "url": "https://www.verint.com/press-room/2026-press-releases/agent-factory-orchestrates-humans-and-ai-workforces/", "title": "Verint Expands Platform With Agent Factory to Orchestrate Human and AI Agent Workforces", "publisher": "Verint", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Announces Agent Factory with capabilities for design/build/orchestrate including human handoffs, prompt management, model flexibility and governance." } ], "claims": [ { "id": "c1", "statement": "Combines prebuilt and custom AI agents, connects them to CX workflows, and routes work to humans when judgment is needed, within one orchestration environment.", "support": "vendor_claim", "sourceIds": [ "s1", "s2" ], "limitations": "Product and press pages; no documentation reviewed on how handoff rules, skills or approval steps are configured.", "capability": "human-handoff-orchestration" }, { "id": "c2", "statement": "Provides centralized controls for managing, updating and governing prompts across agentic AI agents and applications, plus model flexibility including bring-your-own-model.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Announcement-level description; governance mechanics and audit outputs are unspecified.", "capability": "prompt-and-model-governance" }, { "id": "c3", "statement": "Verint announced Agent Factory at its Engage 2026 event as an AI orchestration environment for building, configuring and managing a hybrid workforce of human and AI agents.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Press release does not give a dated general-availability commitment or regional availability.", "capability": "announced-at-engage-2026" } ], "limitations": [ "Evidence is marketing and press material only", "Delegation model is CX-workflow specific" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Included because human-to-agent work routing is the explicit product claim; the CX-only boundary is recorded in the description." }, { "name": "Veza AI Agent Security", "id": "veza-ai-agent-security", "companyName": "Veza", "companyUrl": "https://veza.com", "primaryCategoryId": "identity-access", "secondaryCategoryIds": [ "agent-discovery", "ai-risk" ], "productType": "commercial", "availability": "preview", "buyerRoles": [ "Identity governance lead", "CISO", "Compliance manager" ], "problemsSolved": [ "identifying non-human accounts and their permissions", "right-sizing over-permissioned NHIs", "access reviews covering agents and service accounts" ], "sources": [ { "id": "s1", "url": "https://veza.com/use-cases/nhi-protection-and-governance/", "title": "Non-Human Identity (NHI) Protection and Governance", "publisher": "Veza", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2025-12-07", "supports": "Describes NHI labelling, permission analysis, right-sizing, ownership and lifecycle actions for service accounts and keys." }, { "id": "s2", "url": "https://veza.com/blog/veza-the-enterprise-agent-identity-control-plane/", "title": "Veza - The Enterprise Agent Identity Control Plane", "publisher": "Veza", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": "2026-02-25", "supports": "Announces Veza Access Agents and AI Agent Security including expanded tool discovery and suggested owner mapping; states early access availability." } ], "claims": [ { "id": "c1", "statement": "Veza finds and labels non-human accounts across its integrations, shows their permissions to system resources and supports reviewing them for least privilege, ownership and continued business need.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Use-case page; integration count and label accuracy are vendor-stated.", "capability": "nhi-permission-visibility" }, { "id": "c2", "statement": "Veza identifies the granular tools and actions an AI agent is authorized to invoke within connected applications and maps unmanaged agents and service accounts to suggested human owners.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Blog states these agent capabilities are currently available in early access.", "capability": "agent-tool-entitlement-discovery" }, { "id": "c3", "statement": "Veza Access Agents include an Access Review Agent that prioritizes high-risk items for reviewers with AI-generated recommendations and explanations.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "No evidence of review accuracy or auditor acceptance.", "capability": "ai-assisted-access-review" } ], "limitations": [ "Early-access status limits confidence that agent capabilities are broadly deployable." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Entry scoped to Veza's NHI/agent capabilities, not the whole access-governance suite." }, { "name": "W&B Registry", "id": "wandb-registry", "companyName": "Weights & Biases", "companyUrl": "https://wandb.ai/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "observability" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "ML platform engineer", "MLOps lead", "Head of data science" ], "problemsSolved": [ "scattered model artifacts across teams", "unclear production model provenance", "artifact access control" ], "sources": [ { "id": "s1", "url": "https://docs.wandb.ai/models/registry", "title": "Registry overview — Weights & Biases Documentation", "publisher": "Weights & Biases", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents curated central repository of artifact versions, default Models and Datasets registries, collections, automatic version numbering, linking without duplication, permissions and cross-team sharing." }, { "id": "s2", "url": "https://wandb.ai/site/registry/", "title": "Enhance AI governance with W&B Registry", "publisher": "Weights & Biases", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes versioning, aliases, lineage tracking and governance of models and datasets, lifecycle management from development to staging to production, CI/CD support and reproducibility." } ], "claims": [ { "id": "c1", "statement": "W&B documentation states Registry is a curated central repository of artifact versions within an organization, that organizations start with Models and Datasets registries and can create more, that each registry contains collections representing a task or use case, that artifact versions are assigned automatically (v0, v1, ...) and can be linked into collections without duplication, with permissioned download and cross-team sharing.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs describe artifact curation and access; no approval workflow, risk assessment or control mapping is documented.", "capability": "registry-collections-versioning" }, { "id": "c2", "statement": "The W&B product page describes Registry as providing versioning, aliases, lineage tracking and governance of models and datasets, acting as a single source of truth for which models are in production and enabling reproduction of any model via lineage.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "'Governance' on this page means access control and provenance; it does not evidence policy enforcement or compliance features.", "capability": "governance-positioning" } ], "limitations": [ "Self-hosted/dedicated deployment is offered by W&B generally but was not evidenced on the pages fetched.", "No documented risk, control or regulatory features." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Older W&B material calls this 'Model Registry'; current docs use 'Registry' with a Models registry inside it." }, { "name": "W&B Weave", "id": "wandb-weave", "companyName": "Weights & Biases", "companyUrl": "https://wandb.ai", "primaryCategoryId": "observability", "secondaryCategoryIds": [ "evaluation", "model-governance" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "ML platform lead", "AI engineer", "MLOps manager" ], "problemsSolved": [ "LLM call tracing and cost tracking", "scorer-based evaluation and comparison", "content moderation safeguards", "team-scoped access to AI telemetry" ], "sources": [ { "id": "s1", "url": "https://weave-docs.wandb.ai/", "title": "W&B Weave", "publisher": "Weights & Biases", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes Weave as a framework for tracking, experimenting, evaluating, deploying and improving LLM applications, tracking LLM calls and application logic, automatic tracing and cost tracking, scorers and comparison tools, playground experimentation and pre/post safeguards." }, { "id": "s2", "url": "https://weave-docs.wandb.ai/guides/platform/", "title": "Platform & Security - W&B Weave", "publisher": "Weights & Biases", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents role-based access control at team or project level, SSO with Google, GitHub, Okta and Azure AD via OIDC, SCIM API and SDK, team-based logical separation, restricted scope, scoped service accounts, SOC 2 Type II compliance, HIPAA for Dedicated Cloud, and that BYOB secure storage is not available for Weave." } ], "claims": [ { "id": "c1", "statement": "Weave tracks LLM calls and application logic to debug and analyse production systems, with automatic tracing and cost tracking when connected to existing LLM providers, and supports evaluation with custom or pre-built scorers plus comparison tools.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Docs home does not state capture of tool calls, multi-agent handoffs, sessions or token-level metrics.", "capability": "tracing-and-cost" }, { "id": "c2", "statement": "Platform documentation states role-based access control configurable at team or project level, SSO with public and enterprise identity providers over OIDC, a SCIM API and Python SDK for user and team management, team-based logical separation, a restricted project scope, scoped service accounts, SOC 2 Type II compliance for both platforms and HIPAA compliance for Dedicated Cloud.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Audit logs, retention settings and PII masking are not stated, and the secure storage connector (BYOB) is explicitly unavailable for Weave.", "capability": "access-control-and-compliance" }, { "id": "c3", "statement": "Weave supports pre- and post-safeguards for content moderation and prompt safety.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "No documented policy authoring model, enforcement guarantees or logging of blocked calls.", "capability": "guardrails" } ], "limitations": [ "No documented audit log of user actions or configurable retention for traces.", "Agent-specific telemetry (tool calls, handoffs) is not evidenced on the pages fetched, so agent-governance fit is unproven." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "hybrid" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Access-control and compliance evidence comes from the shared W&B platform page, which explicitly notes one platform feature (BYOB) is excluded for Weave." }, { "name": "Workday Agent System of Record", "id": "workday-agent-system-of-record", "companyName": "Workday", "companyUrl": "https://www.workday.com", "primaryCategoryId": "agent-discovery", "secondaryCategoryIds": [ "work-delegation", "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Workday administrator", "HR technology lead", "AI governance lead" ], "problemsSolved": [ "central catalog of AI agents", "role-based administration of agents", "agent reporting for compliance" ], "sources": [ { "id": "s1", "url": "https://doc.workday.com/admin-guide/en-us/workday-ai/agents/set-up-agent-system-of-record.html", "title": "Set Up Agent System of Record", "publisher": "Workday", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Documents enabling the ASOR functional area, its security domains, per-tenant configuration, and the unconstrained-security-group limitation." }, { "id": "s2", "url": "https://newsroom.workday.com/2025-02-11-The-Next-Generation-of-Workforce-Management-is-Here-Workday-Unveils-New-Agent-System-of-Record", "title": "Workday Unveils New Agent System of Record", "publisher": "Workday", "type": "official_release", "accessedAt": "2026-09-06", "publishedAt": "2025-02-11", "supports": "Announces ASOR as a centralized system for managing a fleet of AI agents from Workday and third parties, with Marketplace-based agent discovery and deployment." } ], "claims": [ { "id": "c1", "statement": "Administrators enable an Agent System of Record functional area and configure domain security policies including Agent Compliance, Agent Management Hub, Manage: Agents, Reports: Agent Reporting and Setup: Agents, and must configure it separately in each tenant.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "Setup documentation describes configuration steps and security domains, not detection capability, agent metadata schema, or third-party onboarding mechanics.", "capability": "agent-administration-domains" }, { "id": "c2", "statement": "Workday announced ASOR on 11 February 2025 as a centralized system for managing an organisation's entire fleet of AI agents from Workday and third parties, with discovery and deployment of agents through Workday Marketplace.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "Announcement-level detail; it does not establish which third-party agent platforms can be registered or how.", "capability": "fleet-management-scope" }, { "id": "c3", "statement": "Documentation states that only unconstrained security groups can currently be configured on ASOR security policies.", "support": "documented", "sourceIds": [ "s1" ], "limitations": "A concrete administrative limitation; its practical impact on delegated administration is not discussed.", "capability": "constrained-security-groups" } ], "limitations": [ "No evidence of detecting unregistered or shadow agents", "Third-party agent registration mechanics not documented in the pages reviewed" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Judgment call: Workday's HCM suite itself is not treated as an agent-governance product (see held file), but ASOR is a distinct, separately configured agent registry with its own security domains, so it is entered here. Its 'digital workforce' framing justifies work-delegation as a secondary category only." }, { "name": "Yields Model Risk Management", "id": "yields-mrm", "companyName": "Yields", "companyUrl": "https://www.yields.io/", "primaryCategoryId": "model-governance", "secondaryCategoryIds": [ "ai-risk" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "Model risk manager", "Model validator", "Chief risk officer" ], "problemsSolved": [ "spreadsheet-based model inventories", "manual validation documentation", "model change control and audit trails" ], "sources": [ { "id": "s1", "url": "https://www.yields.io/solutions/model-risk-management-software", "title": "Model Risk Management Software — Yields", "publisher": "Yields", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes centralized model and AI inventory with classification and risk tiers, workflow engine, controlled model changes with versioning and audit trail, automated MRM documentation and validation reports, and support for independent scalable validation." }, { "id": "s2", "url": "https://www.yields.io/use-cases/ai-inventory", "title": "AI Inventory — Yields.io", "publisher": "Yields", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes structured AI catalogue capturing AI systems, models, agents, use cases, vendors, data sources, deployment context, metadata/classification including EU AI Act risk category, governance metadata and business criticality." } ], "claims": [ { "id": "c1", "statement": "Solutions page states a centralized configurable model and AI inventory with classification and risk tiers, a workflow engine providing standardized auditable workflows across the model lifecycle, controlled model changes with versioning and audit trail, and automatic generation of MRM documentation and validation reports.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Marketing page; no technical documentation was fetched, so automation depth and supported model types are unverified. Approval workflows are not explicitly stated.", "capability": "inventory-workflow-validation-automation" }, { "id": "c2", "statement": "AI inventory page states the catalogue registers AI systems, models, agents, use cases, vendors, data sources (including personal or sensitive data), deployment context, structured metadata and classifications including EU AI Act risk categories, governance metadata such as assumptions and limitations, and business criticality.", "support": "vendor_claim", "sourceIds": [ "s2" ], "limitations": "Field-level scope is vendor-described; capturing an EU AI Act risk category does not establish that the classification is correct or accepted by a regulator.", "capability": "ai-inventory-metadata-scope" } ], "limitations": [ "Product naming varies between 'Yields', 'Chiron' in older material and 'Yields AI Governance Software'; current canonical product name is unclear.", "No public documentation or release notes located." ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas", "self_hosted" ], "geographies": [], "pricing": { "model": null, "url": null }, "formerNames": [ "Chiron" ], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "formerNames records 'Chiron' from historic vendor material; not confirmed on pages fetched this session, flagged in limitations." }, { "name": "Zenity AI Detection and Response (AIDR)", "id": "zenity-aidr", "companyName": "Zenity", "companyUrl": "https://zenity.io/", "primaryCategoryId": "agent-security", "secondaryCategoryIds": [ "observability", "runtime-controls" ], "productType": "commercial", "availability": "generally_available", "buyerRoles": [ "CISO", "SOC manager", "AI security lead" ], "problemsSolved": [ "multi-step agent attack detection", "blocking unauthorised tool calls", "detecting data leakage through agents" ], "sources": [ { "id": "s1", "url": "https://zenity.io/platform/ai-detection-and-response", "title": "AI Detection and Response (AIDR)", "publisher": "Zenity", "type": "official_product", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Describes AIDR as Zenity's runtime security layer for AI agents with sequence-level detection, direct and indirect prompt injection blocking, unauthorised tool invocation blocking, data leakage prevention and agent-to-agent visibility." }, { "id": "s2", "url": "https://zenitydocs.com/", "title": "Zenity documentation", "publisher": "Zenity", "type": "official_docs", "accessedAt": "2026-09-06", "publishedAt": null, "supports": "Lists Zenity Observe, Govern and Defend modules covering observability, AI security posture management and AI detection and response, with intent-breaking detection over full execution paths." } ], "claims": [ { "id": "c1", "statement": "Zenity states AIDR reads the full sequence of chained tool calls, retrievals and actions, detects direct and indirect prompt injection including paraphrased and slow-building attacks, and blocks execution before the agent acts.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "No reference documentation, detection benchmark or false-positive guidance was fetched.", "capability": "sequence-level-detection" }, { "id": "c2", "statement": "AIDR is described as flagging and blocking unauthorised tool invocations or API calls that violate policy and monitoring and blocking sensitive data leaving through agent conversations, tool calls or encoded payloads in real time.", "support": "vendor_claim", "sourceIds": [ "s1" ], "limitations": "Policy authoring model and supported enforcement points are not documented on the page.", "capability": "tool-call-and-data-controls" }, { "id": "c3", "statement": "Zenity's documentation site lists Zenity Observe, Zenity Govern and Zenity Defend covering observability, AI security posture management and AI detection and response.", "support": "documented", "sourceIds": [ "s2" ], "limitations": "The docs landing page does not detail runtime enforcement mechanics.", "capability": "platform-modules" } ], "limitations": [ "Detailed technical documentation not publicly fetched", "Coverage per agent platform not enumerated on fetched pages" ], "reviewStatus": "verified", "reviewedAt": "2026-09-06", "deployment": [ "saas" ], "geographies": [ "IL", "US", "global" ], "pricing": { "model": null, "url": null }, "formerNames": [], "parentCompany": null, "integrations": [], "conflicts": [], "editorialNotes": "Zenity's earlier positioning covered low-code/no-code security; AIDR is the current agent runtime detection product and is recorded on its own." } ]