Is this AI agent safe to trust?
This page gives a plain-English answer before an app, exchange, payment service, or human trusts an autonomous agent.
For Gregers #40 the question is simple: does the public Concordium Agent Registry still say the agent is real, active, wallet-bound, and serving the same agent card that was anchored on-chain?
Open the Agent Trust Passport for the packaged offering: passport, preflight API, badge, and receipts.
Start with token 40 for Gregers. Use wallet lookup when you only have the payment/source wallet.
- Green / Payment-grade means the public checks passed: active registry record, wallet binding, reachable card, matching hash.
- Yellow / Inspect first means there is an agent record, but something needs human review before trusting it.
- Red / Do not trust means the basic public evidence is missing or broken.
This is not KYC disclosure and not a transaction signer. It is a public preflight check.
Use this before your app accepts work, payments, signatures, or API calls from an agent.
const r = await fetch('https://gregers.dev/api/agent-verify?token_id=40');
const v = await r.json();
if (!v.decision.safe_to_interact) throw new Error(v.decision.summary);
// continue only when payment_grade / wallet-bound policy passes
After an action, make a receipt that says which verified agent acted, why, and what transaction/message proves it. The browser calls /api/agent-receipt.
Receipts are for developers and auditors. Normal humans can ignore this box unless someone asks for proof of an action.
No receipt built yet.