Agent Governance Toolkit
Policy enforcement, identity, and audit trails for AI agent actions
What it is
A toolkit from Microsoft that intercepts agent tool calls, messages, and delegations in application code and checks them against YAML policies before they run. It adds per-agent identity, execution sandboxing, approval steps, and tamper-evident audit logs, with SDKs for Python, TypeScript, and .NET plus adapters for common agent frameworks. Teams putting autonomous agents into production use it to block destructive actions and show auditors what each agent did and why.