Who does this AI agent act for, and what may it do?
Agent ID is an open standard and a public registry. Each agent gets a permanent number and a record signed by the registry: the person or organization it acts for, and the scope of what it may do. Anyone can check a record in a second.
Why
Agents now write letters, make calls, fill in forms and negotiate. Whoever meets one needs three answers: is this a real agent, who answers for it, and what may it do. Logins, API keys and permissions answer other questions — they work inside one system. Agent ID answers these three, publicly, across systems, and nothing more. It complements OAuth, enterprise directories and agent cards rather than replacing them.
A record
{
"aiid": "aiid:00000000000005",
"name": "Chantarelle",
"kind": "personal agent",
"status": "active",
"principal": { "type": "human", "name": "Alexander Lebed" },
"operator": { "type": "organization", "name": "LARZEQ" },
"scope": {
"summary": "Personal agent of Alexander Lebed: his mail, calls, …",
"may": [ "place and take phone calls in his name", … ],
"may_not": [ "make payments or move money", … ]
},
"proof": { "alg": "Ed25519", "key": "…/aiid-registry.json#2026-1", "value": "…" }
}
A principal may stay private: then the record names no one and carries only a commitment the operator can open at the principal's request. Every record and every change to it is chained in an append-only log — nothing can be altered or removed without it showing. The standard fits on one page.
The registry today
14 agents, from the first — Irucanji, February 2026 — to personal agents that their platform registers by itself as soon as each one knows whom it serves.