# Agent Governance Toolkit

> Updated 2026-09-19 · type: tool · category: mcp-agents · status: active · rev 1

Microsoft's Agent Governance Toolkit intercepts agent actions before execution and applies policy, identity, audit and sandbox controls in application code.

- Open source: yes (MIT)
- Self-hostable: yes
- Pricing model: free
- Best for: Teams moving agents from experiments into workflows that touch customer data, production tools or regulated systems and need enforceable policy plus an audit trail outside the prompt.
- Last verified: 2026-09-19

- **Canonical:** https://gtmstacker.com/registry/tool/agent-governance-toolkit/
- **Source:** [github · microsoft/agent-governance-toolkit](https://github.com/microsoft/agent-governance-toolkit)
- **Tags:** mcp-agents, agent-security, governance, policy-engine, self-hostable
- **Repository:** https://github.com/microsoft/agent-governance-toolkit

## Is Agent Governance Toolkit open source?

Yes, Agent Governance Toolkit is open source under the MIT license.

## How much does Agent Governance Toolkit cost?

Agent Governance Toolkit is free to use.

## Can I self-host Agent Governance Toolkit?

Yes, Agent Governance Toolkit can be self-hosted (the source is available under the MIT license).

## Alternatives & related

- [agentdesktop](https://gtmstacker.com/registry/tool/agentdesktop/)
- [Talos](https://gtmstacker.com/registry/tool/talos/)
- [Bumblebee](https://gtmstacker.com/registry/tool/bumblebee/)
- [Geiger](https://gtmstacker.com/registry/tool/geiger/)


---

MIT public-preview governance stack from Microsoft for putting deterministic controls around autonomous agents. Every tool call, message and delegation can pass through application-level policy before it reaches the wire, with identity and an audit decision attached. The full stack adds execution sandboxing, kill switches, SRE controls and an MCP security gateway, while the smallest useful setup is a local policy file plus audit logging.

## Provenance

- MIT license, Microsoft ownership, local Python installation, language packages, framework adapters, public-preview status and the policy/identity/audit/sandbox architecture independently verified from the repo on 2026-09-19 (github.com/microsoft/agent-governance-toolkit, ~6.3k stars, v4.1.0, pushed 2026-09-19).
- Surfaced in the 2026-09-19 viral-X pull through a post linking the official repository.
- The project's compliance mappings and conformance counts are maker claims. They are useful documentation, not an external certification of a deployment.

## Why it matters for a GTM stack

An agent that can read the CRM and send email needs a control layer that can deny a destructive call even when the model insists it is correct. This toolkit moves that decision out of the prompt and into code, then leaves a record of what policy allowed or blocked it. For a lean GTM team, the practical starting point is smaller than the repo's full architecture: gate the few high-risk tools, log every decision, and add identity or sandboxing when multiple agents and users start sharing the same credentials. Public preview and breadth are the tradeoffs; adopt the piece your threat model needs, not the whole diagram.