Project
Microsoft's MIT-licensed runtime governance framework for autonomous AI agents, combining a policy engine, mesh, runtime, SRE and compliance modules.
Microsoft's Agent Governance Toolkit README says its policy engine shares a process with the agents it governs, even inside a per-agent container. Zarel, which sells a competing design, argues that governance has to take execution away from the agent entirely.
Reality
Venkat Peri's review of Microsoft's Agent Governance Toolkit finds an LLM call behind its hijacking gate and a 0-1000 trust score deciding agent permissions. Both signals bend under the adversarial pressure they are meant to contain.
Reality
A trade summary of research running through 2025 and into 2026 says agent guardrails fade as sessions grow, and puts the fix outside the model. The one survey figure with a named source is the Cloud Security Alliance's 53%.
Reality
The Linux Foundation now hosts TRACE, a format for evidence of what an AI agent actually did. The package that produces those records still ships as a developer preview.
Reality
Agent Governance Toolkit puts policy checks in the execution path rather than the prompt. The seam: the kernel is middleware inside the agent's process, so containers still do the isolating.
Reality