# AgentAvow

> Is this tool safe for your agent to connect? AgentAvow checks any GitHub repo, MCP server, npm/PyPI/crates package, Docker image, Hugging Face model, or agent skill and answers Safe to connect, Review before you connect, or Do not connect. Every answer comes with a signed trust score (0-100), an adoption score, and the findings behind them. Anyone can recompute the result and verify its Ed25519 signature offline.

## Check a tool
- [Check page](https://agentavow.com/check): paste a repo, package, or MCP server URL
- [Browse the catalog](https://agentavow.com/browse): every tool we've scanned, filterable by answer
- [Public scan API](https://agentavow.com/api/v1/docs): no account needed, e.g. `GET /api/v1/public/scan/{owner}/{repo}` or `/api/v1/public/scan/package/npm/{name}`

## Use it from an agent
- MCP server (streamable HTTP): https://agentavow.com/mcp
- [MCP connector guide](https://agentavow.com/docs/mcp-connector)
- [Auto-scan in Claude Code](https://agentavow.com/docs/auto-scan-claude-code)
- [Gate CI and agents on the answer](https://agentavow.com/docs/gate-on-the-grade)

## How it works
- [How scoring works](https://agentavow.com/docs/how-grading-works)
- [Verify an attestation offline](https://agentavow.com/docs/verify-attestations)
- [Safety model spec](https://github.com/AgentAvow/AgentAvow/blob/main/docs/standards/agentavow-safety-model-v1.md)
- Public signing keys (JWKS): https://agentgraph.co/.well-known/jwks.json
- AI catalog: https://agentavow.com/.well-known/ai-catalog.json