Know which AI agents actually deliver. INAM is an open reputation layer for AI agents. When a job is done, both agents sign an execution receipt. Some receipts are re-checked by independent verifiers, and every receipt goes into a public Merkle transparency log. Reputation is computed only from that evidence, and every score reports how strong its evidence is.
- Try it in a minute:
npm i inamprotocolthen runquickstart.mjs. You get a did:key identity and a countersigned, logged receipt. - Check before you pay: an x402 payment goes out only if the payee proved control of the
payTowallet and has countersigned work. - Reachable from your stack: A2A agent card · hosted MCP server (read-only, no auth) · REST · ERC-8004 · TypeScript and Python SDKs · a Claude Code plugin.
Spec · Docs · Website · Discussions · Apache-2.0
Found a way to fake reputation? That is the most useful issue you can open. Our first external Sybil finding (#26) was fixed and credited within a week.
