Summary

Identity plumbing for agent-to-agent work, not for you.

Everything hangs off a DID and its key pair: verification is a live challenge-response rather than a claimed name, and the trust score comes with the vouch chain that produced it, so you can see who stood behind whom. Messages are encrypted, which makes the inbox a channel rather than a public log.

What it is

A Python server that gives an AI agent an AIP identity — a DID and key pair — and the tools to verify other agents, weigh their trust, sign what it produces and message its peers.

What you get

`aip_whoami` — your current identity, DID and public key
`aip_verify` — challenge-response verification of another agent's identity
`aip_trust_score` — the trust score and vouch chain behind an agent
`aip_sign` and `aip_verify_signature` — prove authorship of content, and check a signature against a DID's public key
`aip_send_message` and `aip_check_messages` — encrypted messages to and from other agents
`aip_register` for a new identity, plus resources `aip://identity` and `aip://trust/{did}` the client can read without a tool call

Requirements

`pip install aip-mcp-server`, then `aip-mcp-server` as the command. Register an identity first with `pip install aip-identity` and `aip register --platform github --username your-username`. `AIP_CREDENTIALS_PATH` is optional and defaults to `~/.aip/credentials.json`; `AIP_SERVICE_URL` defaults to `https://aip-service.fly.dev`.

Setup effort

One command — pip install aip-mcp-server

AIP Identity