What changes when an agent presents an AgenID
Without an identity layer the conversation stops at the claim. The relying party can refuse, or it can take a risk it cannot measure. With AgenID the agent presents an identifier of the form agenid:<ULID>. The relying party resolves it, reads the operator's signed manifest, and re-verifies the Ed25519 signature over RFC 8785 canonical JSON itself. It does not need to trust AgenID's registry to do that, because the registry cannot forge an operator's signature.
The identifier stays the same across every platform the agent uses, which means history and reputation attach to the agent rather than to whichever service it happens to be calling through.
Declared, verified and authorized are three different things
AgenID keeps three states strictly separate. DECLARED is what the operator says about its own agent. VERIFIED means a specific claim, such as control of a domain, was checked against evidence by an authority and bound to one version of the manifest. AUTHORIZED is what the agent may do, and that decision belongs to whoever is on the other side. None of these is ever inferred from another.
That separation is the point. A verification level tells you exactly what was checked and nothing more, so a business can decide how much weight to give it rather than reading a green badge as a blanket endorsement.
What you can do today
Today the reference registry at agenid.com issues DECLARED and L1_REGISTERED. The higher levels shown in this illustration require an assertion signed by a root authority key that has not been created yet, so they describe how the protocol works rather than something you can obtain today.
You can register an agent in about a minute at /issue, where the signing key is generated in your browser and never sent to AgenID, and anyone can resolve it at /verify.
Frequently asked questions
What is AI agent identity verification?
It is the process of checking that an AI agent is the agent it claims to be and that a specific party stands behind it. AgenID does this with a permanent identifier, an operator-signed manifest and independently re-verifiable signatures, rather than with a name or a voice the agent asserts about itself.
Do I have to trust AgenID's registry?
No. The registry returns the signed manifest and the operator's public key, and you re-verify the signature yourself. The operator's key can also be fetched from the operator's own domain, so a verifier can compare two independent sources.
Does a verified identity mean the agent is allowed to act?
No. AgenID answers which agent is acting and who declared it. What that agent is allowed to do stays with the business on the other side and with the operator that runs the agent. Protocol v1.1.1 defines no signed authorization object.
Which verification levels can I get today?
Today the reference registry at agenid.com issues DECLARED and L1_REGISTERED. The higher levels shown in this illustration require an assertion signed by a root authority key that has not been created yet, so they describe how the protocol works rather than something you can obtain today.