CTRL Best AI Agent Governance
CP-04 // REGISTRY, POSTURE & AUDIT LAYER

Observability, Registry & Audit

Control planes for registering enterprise agents, monitoring SaaS and workforce agent usage, and logging action trails. Emphasizes cross-SaaS visibility, posture tracking, and durable audit records of agent activity.

Verified Observability & Audit Control Planes (3)

VendorVerified SummaryCited FactsLatest Citation
Obsidian SecurityForge Global says Obsidian was founded in 2017 by Benjamin Johnson, Matt Wolff, and Glenn Chisholm and is headquartered in Palo Alto, CA.332026-11-21
WitnessAISecurityWeek says California-based WitnessAI was founded in 2023 and has built an AI security and governance platform.332026-11-21
Microsoft Agent 365Microsoft describes Microsoft Agent 365 as an observability platform for AI.212026-11-21

Observability & Audit Dossiers

CP-04

Obsidian Security

Official site · 33 cited facts · Judges A+B passed · 2026-11-21

Verified Overview

Forge Global says Obsidian was founded in 2017 by Benjamin Johnson, Matt Wolff, and Glenn Chisholm and is headquartered in Palo Alto, CA.

Cited Evidence Facts (33)

  1. Forge Global says Obsidian was founded in 2017 by Benjamin Johnson, Matt Wolff, and Glenn Chisholm and is headquartered in Palo Alto, CA. [ forgeglobal.com · 2026-11-21]
  2. SecurityWeek says Obsidian Security offers a platform for governing AI agents and SaaS applications. [ securityweek.com · 2026-11-21]
  3. Forge Global says Obsidian Security is a private company and is not publicly traded. [ forgeglobal.com · 2026-11-21]
  4. Forge Global says Obsidian Security has not yet filed for an IPO. [ forgeglobal.com · 2026-11-21]
  5. EquityZen says Obsidian Security provides a SaaS and AI security platform that helps organizations detect threats, manage security risks, and secure enterprise applications. [ equityzen.com · 2026-11-21]
  6. Obsidian Security says it identifies agents employees run on Codex, Claude Code, Cowork, Cursor, Microsoft Copilot, Amazon Bedrock, Google Vertex, and n8n, and maps each agent to its owner, tools, MCP servers, models, and apps. [ obsidiansecurity.com · 2026-11-21]
  7. Obsidian Security says excessive access and oversharing, along with hardcoded secrets, are detected automatically, and each agent is scored against risk factors when published or changed. [ obsidiansecurity.com · 2026-11-21]
  8. Obsidian Security says its Runtime Policy Decision Engine uses available context on identity, employment, and apps, plus customer data labels, to evaluate agent actions against policy and explain every decision. [ obsidiansecurity.com · 2026-11-21]
  9. Obsidian Security says a dropped table, a mass delete, or customer data sent outside is held for approval or blocked while the agent’s other work continues. [ obsidiansecurity.com · 2026-11-21]
  10. Obsidian Security says a lightweight sidecar intercepts tool calls from Claude Code and Cowork as well as Codex and blocks risky actions before execution. [ obsidiansecurity.com · 2026-11-21]
  11. Obsidian Security says its MCP Security Proxy places a policy checkpoint between agents and MCP tools to permit or block requests before they reach connected apps. [ obsidiansecurity.com · 2026-11-21]
  12. Obsidian Security says its context-aware policy decisions can be integrated into existing gateways through an API. [ obsidiansecurity.com · 2026-11-21]
  13. Obsidian Security says a policy can be written once and applied wherever agents act, with every decision logged for boards and auditors. [ obsidiansecurity.com · 2026-11-21]
  14. Obsidian Security says each decision is logged with the person behind it, the outcome, the reason, and the matching policy, providing a record for auditors when paired with the inventory. [ obsidiansecurity.com · 2026-11-21]
  15. Obsidian Security says Obsidian AI proposes and applies fixes with user approval. [ obsidiansecurity.com · 2026-11-21]
  16. Unite.ai says Obsidian’s Knowledge Graph architecture connects people, accounts, applications, tokens, and entitlements along with their relationships. [ unite.ai · 2026-11-21]
  17. Obsidian Security says the Policy Decision API does not block actions itself: it returns a decision and its reason, which the customer’s gateway, agent harness, or workflow applies. [ obsidiansecurity.com · 2026-11-21]
  18. Obsidian Security says runtime enforcement applies at supported enforcement points: Claude Code and Cowork as well as Codex through a sidecar, Microsoft Copilot Studio through a webhook, supported apps through the MCP Security Proxy, and customer gateways through the Policy Decision API. [ obsidiansecurity.com · 2026-11-21]
  19. Obsidian Security says platforms such as Salesforce Agentforce, Amazon Bedrock, and n8n are covered by discovery and governance. [ obsidiansecurity.com · 2026-11-21]
  20. Obsidian Security says require approval is available where the enforcement point supports it, which currently means Claude Code policies that run before an operation. [ obsidiansecurity.com · 2026-11-21]
  21. Obsidian Security describes its security as including verified controls, measurable compliance, and operational resilience. [ obsidiansecurity.com · 2026-11-21]
  22. Unite.ai says Obsidian describes operating within the user’s existing application permissions rather than adding privileges through a shared service account. [ unite.ai · 2026-11-21]
  23. Obsidian Security describes a financial services company stopping phishing threats and unauthorized AI. [ obsidiansecurity.com · 2026-11-21]
  24. Obsidian Security says BigCommerce enhances SaaS threat detection with Obsidian. [ obsidiansecurity.com · 2026-11-21]
  25. Obsidian Security describes a partnership with T-Mobile. [ obsidiansecurity.com · 2026-11-21]
  26. Obsidian Security says it helped Algolia secure SaaS at scale. [ obsidiansecurity.com · 2026-11-21]
  27. SiliconANGLE says Obsidian Security raised an $85 million Series D funding round at a post-money valuation of $1.1 billion. [ siliconangle.com · 2026-11-21]
  28. Forge Global says Obsidian Security has raised $231.39M to date. [ forgeglobal.com · 2026-11-21]
  29. Forge Global lists Obsidian Security’s major investors as Fama Ventures, Greylock Partners, IVP, Menlo Ventures, Norwest Venture Partners, and Wing VC. [ forgeglobal.com · 2026-11-21]
  30. SecurityWeek says Obsidian will use the funds to accelerate its expansion into agentic AI security. [ securityweek.com · 2026-11-21]
  31. Yahoo says Obsidian expects the funding to support the company until it reaches positive cash flow. [ finance.yahoo.com · 2026-11-21]
  32. Unite.ai says Obsidian Security launched a Runtime Policy Decision Engine designed to let enterprises control AI agent actions inside business applications before execution. [ unite.ai · 2026-11-21]
  33. SecurityWeek says Obsidian’s newest expansion adds native governance controls for Claude Code and Cowork. [ securityweek.com · 2026-11-21]
CP-04

WitnessAI

Official site · 33 cited facts · Judges A+B passed · 2026-11-21

Verified Overview

SecurityWeek says California-based WitnessAI was founded in 2023 and has built an AI security and governance platform.

Cited Evidence Facts (33)

  1. SecurityWeek says California-based WitnessAI was founded in 2023 and has built an AI security and governance platform. [ securityweek.com · 2026-11-21]
  2. WitnessAI lists its address as 2570 W El Camino Real, Suite 150, Mountain View, CA 94040, United States. [ witness.ai · 2026-11-21]
  3. WitnessAI says it brings humans and agents under a unified security and governance platform. [ witness.ai · 2026-11-21]
  4. WitnessAI says it provides visibility into agent activity, control over tool access, and runtime protection for supported agent workflows. [ witness.ai · 2026-11-21]
  5. Axios says WitnessAI helps companies control data flowing into internal AI tools, including how AI agents ingest and move through corporate systems. [ axios.com · 2026-11-21]
  6. PR Newswire says WitnessAI connects human and agentic identities and captures evolving decision-making context, including agent state and execution commands at runtime, to explain agent actions. [ prnewswire.com · 2026-11-21]
  7. PR Newswire says WitnessAI Agentic Control provides visibility and real-time network enforcement over the tools and MCP servers AI agents can access. [ prnewswire.com · 2026-11-21]
  8. WitnessAI says it offers automated red teaming to attack and harden AI models before production deployment. [ witness.ai · 2026-11-21]
  9. WitnessAI says it provides single-tenant isolation, customer-controlled encryption, executive privacy modes, and multi-region deployment to address data sovereignty requirements. [ witness.ai · 2026-11-21]
  10. WitnessAI says it detects and masks sensitive data in real time. [ witness.ai · 2026-11-21]
  11. WitnessAI says it supports meeting industry-specific regulations like PCI DSS 4.0.1. [ witness.ai · 2026-11-21]
  12. WitnessAI says it provides automated security and privacy controls, real-time risk detection, and detailed logging. [ witness.ai · 2026-11-21]
  13. WitnessAI says it routes sensitive requests to approved models based on data classification while maintaining complete records. [ witness.ai · 2026-11-21]
  14. WitnessAI says its bidirectional protection blocks threats such as prompt injection before they reach models and agents, and filters harmful outputs before they reach users or trigger unintended actions. [ witness.ai · 2026-11-21]
  15. PR Newswire says organizations can apply prompt injection, jailbreak, and response protections for AI applications and agents when combined with WitnessAI's AI Firewall capabilities. [ prnewswire.com · 2026-11-21]
  16. WitnessAI says its data protection tokenizes sensitive information, its model protection defends against manipulation, and its agent guardrails enforce rules of engagement for autonomous systems. [ witness.ai · 2026-11-21]
  17. WitnessAI says it supports an organization-wide approved-tool list of MCP servers and tools, enforced at the network for every agent across every IDE, chat app, and custom-built agent. [ witness.ai · 2026-11-21]
  18. WitnessAI says organization-level bans cannot be re-enabled by a team admin or routed around by an agent switching providers. [ witness.ai · 2026-11-21]
  19. WitnessAI says every blocked tool call generates an audit record identifying the user, agent, tool, and rule. [ witness.ai · 2026-11-21]
  20. WitnessAI says it traces every agent action to a human identity for audit. [ witness.ai · 2026-11-21]
  21. WitnessAI identifies InComm Payments as using WitnessAI for safe AI usage. [ witness.ai · 2026-11-21]
  22. SecurityWeek says WitnessAI reports use of its platform by large enterprises in the airline, automobile, financial services, retail, telecommunication, and utilities sectors. [ securityweek.com · 2026-11-21]
  23. PR Newswire says WitnessAI raised $27.5 million in a Series A funding round in May 2024, co-led by Google Ventures and Ballistic Ventures. [ prnewswire.com · 2026-11-21]
  24. TechCompanyNews says WitnessAI secured $58 million in a strategic funding round led by Sound Ventures, with participation from Fin Capital, Qualcomm Ventures, Samsung Ventures, Forgepoint Capital Partners, and Artisanal Ventures. [ techcompanynews.com · 2026-11-21]
  25. The Los Angeles Times says the funding will be used to accelerate WitnessAI's global go-to-market and product expansion. [ latimes.com · 2026-11-21]
  26. The Los Angeles Times says WitnessAI reported more than 500% growth in annual recurring revenue in the past 12 months. [ latimes.com · 2026-11-21]
  27. The Los Angeles Times says WitnessAI reported scaling employee headcount by five times in the past 12 months. [ latimes.com · 2026-11-21]
  28. PR Newswire says the WitnessAI platform delivers a 99.3% true positive rate on employee AI guardrails, extending the same platform foundation that governs AI usage across more than 4,000 AI applications and over 100 supported model types. [ prnewswire.com · 2026-11-21]
  29. PR Newswire says WitnessAI Agentic Security is available January 2026. [ prnewswire.com · 2026-11-21]
  30. TechCompanyNews says WitnessAI Agentic Security monitors agent activities, connections, and data flows while blocking malicious prompts based on behavioral intent. [ techcompanynews.com · 2026-11-21]
  31. PR Newswire says WitnessAI Agentic Control is available to all customers. [ prnewswire.com · 2026-11-21]
  32. PR Newswire says WitnessAI's new MCP Catalog scores known tools against OWASP and CVE risk classes so security teams can assess risk before granting approval. [ prnewswire.com · 2026-11-21]
  33. WitnessAI says it was named a founding member of the ServiceNow Control Tower Alliance. [ witness.ai · 2026-11-21]
CP-04

Microsoft Agent 365

Official site · 21 cited facts · Judges A+B passed · 2026-11-21

Verified Overview

Microsoft describes Microsoft Agent 365 as an observability platform for AI.

Cited Evidence Facts (21)

  1. Microsoft describes Microsoft Agent 365 as an observability platform for AI. [ learn.microsoft.com · 2026-11-21]
  2. bismart.com says Microsoft describes Agent 365 as a centralized platform for deploying, orchestrating and governing agents built with Microsoft technologies, open-source frameworks or third-party solutions. [ blog.bismart.com · 2026-11-21]
  3. ic-consult.com says Microsoft Agent 365 provides a central agent registry that discovers agents built in Copilot Studio and Microsoft Foundry, along with agents from third-party platforms and custom frameworks. [ ic-consult.com · 2026-11-21]
  4. Microsoft says Agent Map displays agents on a visual map grouped by the platform on which they were built. [ learn.microsoft.com · 2026-11-21]
  5. Microsoft says cost management shows organizational Copilot Credits usage broken down by user, group, agent and service. [ learn.microsoft.com · 2026-11-21]
  6. Microsoft says users can set spending limits for groups and users and see who is approaching those limits. [ learn.microsoft.com · 2026-11-21]
  7. Microsoft says users review and approve agents before publishing them to their organization and decide who can use them. [ learn.microsoft.com · 2026-11-21]
  8. Microsoft says users control which tools agents can connect to and can apply a standard set of policies to each new agent. [ learn.microsoft.com · 2026-11-21]
  9. VentureBeat says Agent 365 provides an SDK with varying levels of integration for third-party agent frameworks, including LangChain, CrewAI and other open-source tools. [ venturebeat.com · 2026-11-21]
  10. Microsoft says policies can apply to agent identities, target agent resources and trigger based on agent risk. [ learn.microsoft.com · 2026-11-21]
  11. Microsoft says web content filtering, threat intelligence filtering and network file filtering can be applied to agent traffic. [ learn.microsoft.com · 2026-11-21]
  12. Microsoft says users can immediately block an agent that looks risky while their security team investigates. [ learn.microsoft.com · 2026-11-21]
  13. Microsoft says Microsoft Defender watches for threats and blocks unsafe or harmful activity. [ learn.microsoft.com · 2026-11-21]
  14. Microsoft says Microsoft Agent 365 is licensed on a per-user basis. [ learn.microsoft.com · 2026-11-21]
  15. reply.com says Agent 365 is available standalone at $15 per user per month. [ reply.com · 2026-11-21]
  16. ic-consult.com says Agent 365 is available as a standalone license or as part of Microsoft 365 E7. [ ic-consult.com · 2026-11-21]
  17. KPMG says it will adopt Microsoft Agent 365 to manage how AI agents are deployed, managed, monitored and updated across the KPMG global organization. [ kpmg.com · 2026-11-21]
  18. KPMG says that, under the agreement, it will use Microsoft Agent 365 to enhance the KPMG Trusted AI framework and help clients manage, monitor and secure AI agents across their organizations. [ kpmg.com · 2026-11-21]
  19. VentureBeat says Avanade offered an early endorsement of Agent 365. [ venturebeat.com · 2026-11-21]
  20. Microsoft says Microsoft Agent 365 is generally available for the Commercial segment on a per user basis. [ learn.microsoft.com · 2026-11-21]
  21. Microsoft says at least one user must be licensed with a qualifying Microsoft Agent 365 license to enable Agent 365. [ learn.microsoft.com · 2026-11-21]

Observability & Audit Sources (28)

  1. https://www.obsidiansecurity.com/ai-agent/ai-agent-security [obsidiansecurity.com · 2026-11-21 · 11 fact s]
  2. https://learn.microsoft.com/en-us/microsoft-agent-365/overview [learn.microsoft.com · 2026-11-21 · 10 fact s]
  3. https://witness.ai/product/ [witness.ai · 2026-11-21 · 7 fact s]
  4. https://witness.ai [witness.ai · 2026-11-21 · 6 fact s]
  5. https://forgeglobal.com/obsidian_ipo/ [forgeglobal.com · 2026-11-21 · 5 fact s]
  6. https://www.prnewswire.com/news-releases/witnessai-introduces-agentic-control-to-secure-and-govern-ai-agents-and-mcp-servers-302802606.html [prnewswire.com · 2026-11-21 · 5 fact s]
  7. https://www.obsidiansecurity.com/customers [obsidiansecurity.com · 2026-11-21 · 4 fact s]
  8. https://witness.ai/for-compliance/ [witness.ai · 2026-11-21 · 3 fact s]
  9. https://www.latimes.com:443/b2b/ai-technology/story/2026-02-09/witnessai-raises-58m-sound-ventures [latimes.com · 2026-11-21 · 3 fact s]
  10. https://www.obsidiansecurity.com [obsidiansecurity.com · 2026-11-21 · 3 fact s]
  11. https://www.prnewswire.com/news-releases/witnessai-raises-58-million-for-global-expansion-and-announces-new-ways-to-secure-ai-agents-302659319.html [prnewswire.com · 2026-11-21 · 3 fact s]
  12. https://www.securityweek.com/obsidian-security-raises-85-million-at-1-1-billion-valuation/ [securityweek.com · 2026-11-21 · 3 fact s]
  13. https://www.unite.ai/obsidian-security-runtime-policy-engine-ai-agent-actions/ [unite.ai · 2026-11-21 · 3 fact s]
  14. https://ic-consult.com/en/resources/blogs/ai-agent-governance-with-agent-365-and-entra-agent-id-discover-govern-and-secure-the-agents-in-your-tenant/ [ic-consult.com · 2026-11-21 · 2 fact s]
  15. https://kpmg.com/xx/en/media/press-releases/2026/06/kpmg-and-microsoft-scale-trusted-enterprise-ai-agents-globally.html [kpmg.com · 2026-11-21 · 2 fact s]
  16. https://learn.microsoft.com/en-us/microsoft-agent-365/admin/capabilities-entra [learn.microsoft.com · 2026-11-21 · 2 fact s]
  17. https://venturebeat.com/technology/microsoft-says-ungoverned-ai-agents-could-become-corporate-double-agents-its [venturebeat.com · 2026-11-21 · 2 fact s]
  18. https://www.securityweek.com/witnessai-raises-58-million-for-ai-security-platform/ [securityweek.com · 2026-11-21 · 2 fact s]
  19. https://www.techcompanynews.com/witnessai-raises-58-million-in-funding/ [techcompanynews.com · 2026-11-21 · 2 fact s]
  20. https://blog.bismart.com/en/microsoft-agent-365-ai-agent-governance [blog.bismart.com · 2026-11-21 · 1 fact ]
  21. https://equityzen.com/company/obsidiansecurity/ [equityzen.com · 2026-11-21 · 1 fact ]
  22. https://finance.yahoo.com/technology/ai/articles/obsidian-security-raises-funding-1-120313720.html [finance.yahoo.com · 2026-11-21 · 1 fact ]
  23. https://learn.microsoft.com/en-us/microsoft-agent-365/ [learn.microsoft.com · 2026-11-21 · 1 fact ]
  24. https://siliconangle.com/2026/08/04/obsidian-security-raises-85m-ai-agents-create-cybersecuritys-next-major-attack-surface/ [siliconangle.com · 2026-11-21 · 1 fact ]
  25. https://witness.ai/resources/?cat=case-studies [witness.ai · 2026-11-21 · 1 fact ]
  26. https://www.axios.com/2026/01/13/witnessai-funding-enterprise-ai [axios.com · 2026-11-21 · 1 fact ]
  27. https://www.obsidiansecurity.com/trust [obsidiansecurity.com · 2026-11-21 · 1 fact ]
  28. https://www.reply.com/valorem-reply/en/resources/insights/blog/azure/what-is-microsoft-agent-365-the-control-plane-for-enterprise-ai-agents-explained [reply.com · 2026-11-21 · 1 fact ]