Traceable AgentsA public register of AI agents

Look up an AI agent.

Every agent on this register has an ID, an owner who signed for it, and a record of what it has done. Paste an agent's ID to see who stands behind it and whether its record is clean.

You get one of three answers: identified with a verified owner, identified with an owner on record, or unknown.

How it works

  1. An owner signs for the agent. The owner is a company or a person. They hold a root key and sign each agent's ID with it. An ID with no owner's signature counts as unknown.
  2. Every action adds a receipt. Each time the agent acts through a verifier, a signed receipt joins its record: what kind of action, what the verdict was, and when. The content of the action stays with the owner. The register holds only hashes.
  3. Anyone can check. Before your systems deal with an agent, look it up. The answer, the owner's signature and the whole receipt chain can be checked from the public keys alone. You do not have to trust this site.

What the answer means

Identified, owner verified
A named owner signed for this agent and the register has confirmed who they are.
Identified, owner on record
An owner signed for this agent. The register has not yet confirmed who they are.
Unknown
No agent with this ID is on the register. Treat it as anonymous.

What an agent may do with each answer is your decision. A bank might let an unknown agent read a price list and nothing more.

Get your agents on the register

Any system that runs or checks AI agents can issue IDs and write receipts in the published format. The first one is Sheaf: every agent that runs through Sheaf gets an ID under its owner's key, and every checked action adds a receipt. How to join as a verifier.