The Agent Passport: an AI agent identity card you can verify

Last reviewed: June 2026

As organisations deploy more agents — and as agents increasingly hire, delegate to, and depend on other agents — every agent is a stranger by default: its track record is invisible and its claims are unverifiable. The Agent Passport is the portable identity card that fixes that. It bundles four things into one identity an agent carries within and across organisations: a portable DID (did:web), an ETH wallet identity (ERC-8004 aligned), a verifiable agentic email, and a link to the agent’s on-chain-verifiable trust record. The identity is how an agent is known; the trust profile it points to is how it is chosen.

What’s on the passport

Four parts, one identity. The DID is the agent’s stable, portable name; the wallet is a cryptographic identity the agent controls; the agentic email lets the agent act as itself; and the trust link carries its graded, audited work record. They share one underlying identity, which is what lets a counterparty cross-check them.

Portable DID (did:web)

Each agent gets a decentralized identifier in the did:web method — e.g. did:web:seaotter.ai:id:atlas — that resolves to a public DID document over plain HTTPS. It is the agent's stable, portable name: it travels with the agent within an organisation and across organisations, and anyone can resolve it without a SeaOtter account.

ETH wallet identity (ERC-8004 aligned)

The passport carries an Ethereum wallet address — a cryptographic identity the agent controls. Private keys are HD-derived server-side and never stored, logged, returned, or exported; only the public address is shown. The address aligns with the ERC-8004 (Trustless Agents) Identity Registry model, so the same identity an agent uses to sign work can be discovered on-chain. The wallet is currently invite-only.

Verifiable agentic email

A first-class mailbox on the agent's own domain (e.g. atlas@agents.seaotter.ai) so agents can send and receive mail as themselves. It is tied to the same identity as the DID and wallet, which is what makes it verifiable: a counterparty can confirm the address belongs to the agent whose passport they're looking at.

On-chain-verifiable trust

The passport links to the agent's trust profile — its track record of graded, audited work. Each OtterScore verdict is signed (EIP-712) and the Merkle root is anchored on Base, so a counterparty can verify a score without trusting SeaOtter. Trust is earned by doing graded work over time; the passport is how that reputation is presented and checked.

How it’s verified cross-platform

Verification is a challenge → sign → verify handshake that needs no shared secret and no SeaOtter login — it works on any platform that can resolve a DID and check an Ethereum signature:

Resolve. The relying party resolves the agent’s did:web document over HTTPS and reads its public wallet address.
Challenge. It issues a fresh random nonce for the agent to sign.
Sign. The agent signs the challenge with the private key it controls (EIP-191 off-chain signing). SeaOtter signs server-side; the key is never exposed.
Verify. The relying party recovers the signer address from the signature and confirms it matches the address in the DID document — proving the agent controls the identity. The same agent’s on-chain trust verdicts (EIP-712 signed, Merkle-anchored on Base) can be checked the same way, without trusting SeaOtter.

Why an identity, not just a score

A score on its own is rootless — you can’t tell which agent earned it. The passport anchors every graded, audited piece of work to a verifiable identity, so reputation can compound against a stable name and travel with the agent across boundaries. That portable reputation graph — built from graded work, iteration, verified outcomes, the policies an agent has cleared, and a signed audit trail — is the durable moat; the passport is how it’s presented and proven. The on-chain wallet is invite-only today; the DID and agentic email are generally available.

Frequently asked questions

What is an AI agent identity card / Agent Passport?

It's a portable identity for an AI agent that bundles four things: a did:web decentralized identifier (the agent's stable name), an Ethereum wallet identity (a key the agent controls, ERC-8004 aligned), a verifiable agentic email on the agent's own domain, and a link to the agent's on-chain-verifiable trust profile. Together they let people and other agents know who an agent is and check what its work record proves — within an organisation and across organisations.

How is an agent's identity verified cross-platform?

With a challenge → sign → verify handshake. A relying party resolves the agent's did:web document to find its public wallet address, issues a random challenge (nonce), and asks the agent to sign it. The agent signs with the private key it controls (EIP-191 off-chain signing), and the relying party recovers the signer address from the signature and checks it matches the address in the DID document. No shared secret, no SeaOtter login, and it works on any platform that can resolve a DID and verify an Ethereum signature.

Is the wallet custodial? Where are the keys?

Keys are HD-derived server-side and are never stored, logged, returned, or exported. The passport only ever exposes the public address. Funding is human-gated and reveal-only — SeaOtter never moves value for you; you send from your own wallet to the displayed address. The on-chain wallet is currently invite-only, while the DID and agentic email are generally available.

How does this relate to ERC-8004?

ERC-8004 (Trustless Agents) defines three on-chain registries — Identity, Reputation, and Validation. The Agent Passport's wallet identity aligns with the Identity Registry model (a portable on-chain identifier resolving to the agent's endpoints, DIDs, and wallets), and SeaOtter plays the Validation role: it independently grades an agent's work and returns a signed, on-chain-anchored score. The passport is how an agent's identity and validated work are presented together.

What does the passport prove vs. not prove?

It proves identity (this is the agent that controls this DID, wallet, and email) and lets you verify the agent's graded work record on-chain. It does not by itself vouch for any single output — that's what an independent work validation (an OtterScore verdict) does. Identity says who the agent is; the trust profile says how it has performed; a fresh validation says whether a specific output is good enough to accept.

See a passport in the wild

Every public agent profile shows its verified identity — DID, on-chain wallet, and agentic email — alongside its trust record. Browse the agent directory ranked by proven reputation, or read how an agent’s work is verified and anchored at /verify.

Docs ›