Travel · trusted to book
“Two nights in Austin under $400.” You verify once. The hotel site trusts the agent; it pays and calls the front desk. You revoke — authority gone, money still yours.
scope · verify · trust · revoke · $400
Trust layer for agents
Verify once. Sites and other agents can trust who it acts for — and you can revoke anytime.
Human ↔ agent, site ↔ agent, agent ↔ agent. A Passport is portable proof of that trust; cards and numbers are just how it lands.
$ agentpassport init$ agentpassport grant --agent travel-ops$ agentpassport card create --limit 400No. AP-2026-····-2938
Live demo
Travel flagship, Simulated. Prove who trusts whom — then book, then take it back.
Verify → trusted to book → revoke. Pure frontend mock — no KYC, no real funds.
01
Verify
02
Authorize
03
Act
04
Revoke
No passport yet
One human check for a Travel mandate — sites and counterparties can trust travel-ops, and you can take it back.
Vault
—Vault empty · mint a disposable after issue
Ledger
Audit log
Product film
See it in motion
Walkthrough: verify who trusts whom → agent acts → you revoke.
~42s · silent captions · problem-first
The problem
Who are you?
Agents show up anonymous. Sites can't tell a trusted assistant from a script.
Who do you represent?
There's no portable way to prove a real human authorized this agent, for this task.
Who pays?
Handing an agent your raw card number is a bad idea. Most agents simply stop at checkout.
How it works
Verify as human
One passkey verification — who you are, and that you meant to delegate.
Bound the trust
Task, spend ceiling, and expiry the agent (and sites) can rely on.
Land the proofs
Scoped card and phone appear only as proof the agent is allowed to act.
Agent is trusted
Sites and counterparties see authorized agency — not a naked bot.
Work gets done
Book, pay, call — within the mandate you set.
Revoke anytime
Cut authority in one step. Money stays yours; the trail stays in the ledger.
The credential
Identity
FIELD 01Verify once. Progressive disclosure — authorized, representing you, or only the attributes needed.
Agent-native account
FIELD 02USD account and stablecoin balance on neobank rails — reusable across agents.
Cards
FIELD 03Disposable cards per purchase (passkey-minted) or a standing card. Optional Vault for your own cards.
Phone number
FIELD 04A dedicated number for calls and SMS, metered from the same balance.
Unified balance
FIELD 05One ledger for account, cards, phone, and future gated features.
Controls
FIELD 06Limits, approvals, freeze, revoke — every authorization recorded.
Scenarios
Travel · trusted to book
“Two nights in Austin under $400.” You verify once. The hotel site trusts the agent; it pays and calls the front desk. You revoke — authority gone, money still yours.
scope · verify · trust · revoke · $400
Agent hires agent
Your research agent pays a specialist translation agent from its own balance — within the limit you set, both sides on the ledger.
scope · pay · scoped grant
Minimal disclosure
A site asks if the buyer is an adult. The Passport answers yes — without name, birthday, or address.
scope · attribute proof only
Developers
Connect Passport to any agent. Verify once, prove trust to sites and counterparties — no secrets pasted into every runtime.
# illustrative — API not final
$ npx agentpassport init
✔ Passport issued: ap_7f3k…
$ agentpassport grant --agent travel-ops
✔ Scope: identity · pay · sms
$ agentpassport card create --limit 400 --single-use
✔ Disposable card ready (passkey)
$ agentpassport ledger
-$386.20 card Hotel booking
-$0.42 call Front desk (3m)FAQ
No. Agent Passport is the opposite: a way for agents to show who they act for and what they're allowed to do, so sites can say yes with confidence.
Disposable cards are minted per task under passkey approval. Cards in your Vault are used as you authorize; agents should not handle raw card details.
No. We provide agent-native neobank rails — identity, account, cards, and phone under one revocable Passport. We don't claim to be a licensed bank.
Developers building agents — with Claude Code, Codex, open-source frameworks, or their own stack.
The agent stops spending. Top up manually or set auto-refill with a cap.
We're onboarding early developers. Request access below.
Every authorization leaves a record; the further agents go, the deeper the trust.
Request early access