
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
Encrypted vault for .env files with scoped, audited MCP access for coding agents
Passive audit that checks whether an MCP endpoint exposes its tools without authentication
Runs locallyPay-per-call audit of a public URL: redirects, TLS, security headers and DNS records
Free read-only verification tools for agents, with a signed receipt on every call
Score a domain's DNS, email authentication and TLS posture, with fixes

Open-source control plane that firewalls MCP tool calls and caps agent model spend
Check an AI agent's action against signed rules before it commits, with auditable proof
Sealed microVM sandboxes for agents to run untrusted code, locally or in the cloud
Not checked yetOne call to verify agent output, price its resource use and enforce spending mandates
Not checked yetRecord AI agent actions as hash-only, tamper-evident audit evidence
Runs locallyRead-only directory of GRC and security AI agents and verified audit firms
Not checked yetList AugmentEV's agents; API key required and accounts are by request
Not checked yetGrade your installed MCP servers A to D from public signals, with safer alternatives
Runs locally
Hash and sign evidence of what AI agents did, mapped to AI governance frameworks
Runs locally
Action catalog and approval policy that let agents run server tasks without SSH
Not checked yet
Read-only MCP gateway that approves exact tools and blocks changed or unknown calls
Not checked yetSpend checks for AI agents: duplicate invoices, shared budgets and signed receipts
Not checked yetAuthorize AI agent actions before they run, with policy, approval and signed grants
Not checked yetStamp tool results with their context and reject stale, duplicate or wrong-task outputs
Not checked yet
Free source-linked evidence on MCP and agent tool trust, plus optional paid local checkers
Not checked yetMCP server giving agents a signed identity and hash-anchored settlements on Midnight
Runs locallyPre-payment check for AI agents: allow, hold or block, with a signed attestation
Not checked yetRead-only MCP endpoint that measures an MCP server or A2A agent against five conditions
Not checked yetHand AI agents scoped API keys from an encrypted vault, with an audit trail
Not checked yetRegistry, scoped permissions and one-step recall for AI agents in health systems
Not checked yetAudits tool-name collisions before a new MCP server joins your merged toolbox
Not checked yetFact-check LLM output claim by claim and gate risky agent actions with a safety layer
Not checked yetCheck whether a public MCP server's declared tools have changed since last seen
Not checked yet