A passport for every AI agent.
Verifiable identity for AI agents, separate from the people and companies behind them. It proves who is acting — and anyone checks it offline, in a few lines of code.
The agent holds its own key logged before it is valid revoke once, every running copy is refused
Read https://ainra.org/skills.md and verify an AINRA passport.One line for your agent. A few for your code.
No account, no calls to us — signed facts, checked locally. Nothing to install? Run the verifier in this browser.
Verifying a running copy rather than the agent itself? It carries a short-lived credential instead of the lineage’s key — you supply your own audience and the same call checks it. The instance path.
One command. Your coding assistant gains AINRA's skills — verify, issue, present, renew, revoke.
npx skills add https://ainra.org/.well-known/skills/index.json
window.ainra.selfTest() · run it with nothing installed ↗
One root. Three rings.
Independent registrars issue, agents carry passports, verifiers decide — three rings, one neutral centre.
Keys, rules, revocation, logs. Everything chains up to it; every check stays local and fast.
Separate organizations issuing under identical published rules.
Each carries a passport chained to the root; its running copies get short-lived keys.
Anyone deciding whether to trust an agent — the check runs on their own machine, and each picks its own bar.
Accredit · anchor · revoke · log — the root does four jobs and nothing else. It returns facts, never scores; every check runs on the verifier's own machine, and the decision stays theirs.
Soon, your AI will act for you.
Some already act alone: intrusion campaigns run, shutdowns evaded — all publicly reported. Every response presupposes what they lack: an identity that cannot be shed.
Anything can claim to be your agent.
Nobody can tell a real one from a fake.
A passport it can prove.
Who it is, what it may do — signed. Checked in milliseconds; revoked in under a minute.
Yours, not ours.
You set the limits, you pull the switch — we just make them provable.
- Who created this agent?
- Who controls it now?
- What is it allowed to do?
- Has it been verified — or revoked?
- Is it pretending to be another?
- Can it legally act for a company?
- Can I safely delegate to it?
Tomorrow's checks look like today's — KYC for people, KYB for companies, KYA for agents: every acting AI carrying a scannable ID with a machine-readable zone.
Watch the case for urgency.
“A national security emergency” — Wyatt Tessari L’Allié of AI Governance and Safety Canada, on agent loss-of-control. Their argument stops at the alarm; ours starts there: every response needs identity, and identity needs a neutral root.
AINRA is a proposal for that neutral root — a non-profit standard anyone can run, own, or fork. You are not asked to trust its founders; everything is public.
Status. Currently pre-ceremony: an operator-run genesis root exists; the recorded public ceremony across independent custodians is the pending milestone. Three implementations agree on all 1153 CC0 conformance vectors, on verdict and reason, and a fourth agrees on canonical encoding; revocation is measured under 60 seconds in a local drill; verification is offline-capable, free, and phones home to no one. The passport above is a labeled specimen. By charter, enforced in code, the root issues no passports, computes no scores, holds no personal data, and sells nothing. Canonical live status: foundation.html.