
Microsoft toolkit for policy enforcement, identity and audit of autonomous agents
- Open source
- Open-source agent
- 6.6K GitHub stars
- Security

Microsoft toolkit for policy enforcement, identity and audit of autonomous agents

Open-source gateway adding identity, policy and audit between AI agents and your systems

Audits MCP server configs for leaked secrets, shell injection and unpinned package runners

Reviews a proposed agent action or approval packet against its evidence before execution

Audits an AI agent codebase against the OWASP ASI Top 10 agentic security risks

Applies seven rules of thumb to accept, dismiss or query vulnerability reports
Postgres gateway for AI agents with table grants, column masking and approval for writes
Runs locallyRead-only audit of agent skills in one workspace, plus a package manager for them
Runs locallyPassive audit that checks whether an MCP endpoint exposes its tools without authentication
Runs locallyA lockfile for MCP tools: blocks changed tool definitions and signs a receipt per call
Runs locallyBrowser actions for AI agents that are classified, approved and logged
Runs locallyRead-only Trino access for agents, run as the verified user so Trino policy decides
Runs locallyOpen-source MCP gateway that pins tools, holds risky calls and limits egress
Runs locally
Open-source control plane that firewalls MCP tool calls and caps agent model spend
Policy-gated agent spending with signed receipts and audit against the payment rail
Runs locally
Offline AI-law rule checks for the EU AI Act, NYC LL 144, GDPR and HIPAA
Runs locallyOpen-source scheduler that runs coding agents in parallel under policy-as-code rules
Runs locallyAudit remote MCP servers for OAuth 2.1 gaps, and generate tested connectors
Runs locally
Hash and sign evidence of what AI agents did, mapped to AI governance frameworks
Runs locallyProxmox VE access across many clusters, with deletes held behind plan, approve and apply
Runs locallyRegistry that records who authorised an AI agent, within what limits and until when
Runs locallyRust MCP server for the Juniper Mist cloud with scoped tokens and two-person changes
Runs locallyRust MCP server for UniFi controllers with scoped tools and two-person change control
Runs locallyRun MCP servers with a scoped environment, audit trail and discovery from registries
Runs locallyOperate Proxmox through an AI agent with planned changes, receipts and undo
Runs locally
File ownership, expiring shares and audit trail for SaaS apps, exposed to agents over MCP
Runs locallyOpen-source gateway that routes AI requests by privacy policy and governs tool access
Not checked yetPassive scan of a website's AI-agent readiness: robots, llms.txt, JSON-LD, MCP, security
Runs locallySelf-hosted gateway that masks PII and signs receipts for AI access to your databases
Runs locallyRecord AI agent verdicts bound to file hashes and refuse stale ones before shipping
Runs locally