# Agent Transaction Passport™ (ATP) — Draft Specification v0.1

**Published by AgentProof**
**Status:** Public draft
**Category:** Autonomous Transaction Trust Infrastructure
**Canonical URL:** https://agentproof.hatchable.site/standards/agent-transaction-passport-v0.1.md
**Machine schema:** https://agentproof.hatchable.site/standards/agent-transaction-passport-v0.1.schema.json

## Purpose

An Agent Transaction Passport™ (ATP) is a portable, machine-verifiable record of an economically or operationally meaningful transaction involving one or more autonomous agents.

The passport does not replace payment rails, identity systems, agent protocols, contracts, or credentials. It binds their outputs into a durable transaction record so another party can answer:

- who or what acted;
- for whom the agent acted;
- what authority the agent had;
- what the parties intended and agreed;
- what evidence was relied upon;
- what was delivered;
- how settlement occurred;
- what attestations were made;
- and whether the record has been modified.

## Core vocabulary

**Agent Transaction Passport™ (ATP)** — the complete portable record for one agent-mediated transaction.

**Principal** — the human, organization, software entity, or other accountable party on whose behalf an agent acts.

**Agent** — the autonomous or semi-autonomous software actor participating in the transaction.

**Mandate** — the authority granted to an agent, including scope, limits, expiration, and source credentials.

**Counterparty** — the entity or agent on the other side of the transaction.

**Evidence Set** — the source material, observations, hashes, receipts, and attestations relied upon before or during the transaction.

**Agreement Record** — the machine-readable record of what the parties committed to exchange or perform.

**Fulfillment Record** — evidence of what was actually delivered or performed.

**Settlement Record** — proof of payment or other value transfer, independent of the underlying payment rail.

**Attestation** — a signed or attributable statement about identity, authority, evidence, fulfillment, or settlement.

**Integrity Digest** — the cryptographic digest of the canonical passport payload.

**Passport Chain** — an optional hash link from a passport to an earlier related passport, enabling auditable sequences without requiring a blockchain.

## Design principles

1. **Rail-neutral.** ATP can reference x402, AP2, MPP, cards, bank transfers, stablecoins, internal ledgers, or future settlement systems.
2. **Protocol-neutral.** ATP can reference MCP, A2A, vendor-native agent protocols, or direct HTTP interactions.
3. **Evidence-first.** Assertions should point to independently retrievable evidence or cryptographic digests whenever possible.
4. **Selective disclosure.** Implementations may disclose hashes, credentials, or derived claims instead of confidential payloads.
5. **Portable.** A verifier should not need the original chat session or orchestration environment.
6. **Tamper-evident.** Canonical payloads should be hashed and, where available, signed.
7. **Extensible.** New credential, payment, jurisdiction, and attestation types can be added without redefining the base object.

## Minimum passport

A conforming v0.1 passport MUST contain:

- `spec`
- `passport_id`
- `issued_at`
- `issuer`
- `principal`
- `agent`
- `mandate`
- `counterparty`
- `agreement`
- `evidence`
- `fulfillment`
- `settlement`
- `attestations`
- `integrity_sha256`

## Verification

A verifier SHOULD:

1. validate the object against the published schema;
2. recompute the canonical payload digest;
3. verify signatures or credentials referenced by attestations;
4. verify referenced payment/settlement records when possible;
5. verify evidence URLs, content hashes, or external receipts where available;
6. evaluate mandate scope and expiration;
7. treat unverifiable fields as claims, not facts.

## AgentProof authorization binding extension

A passport MAY include an `authorization` object that binds the completed transaction record to a previously registered AgentProof trust-resolution decision.

The reference extension records the resolution ID, decision, policy version, protocol version, resolution SHA-256, and—when recursive authority was used—the `authority_chain_sha256`.

The hosted ATP issuer only binds an AgentProof resolution when the stored decision is `ALLOW` or `ALLOW_WITH_LIMITS` and the stored principal, agent, and counterparty fingerprints match the passport request.

This creates a durable bridge between **pre-transaction admission** and the **post-transaction passport** without making authorization a mandatory ATP field for historical or externally authorized transactions.

## Relationship to AgentProof

AgentProof is an implementation and verification service for the ATP standard. Existing AgentProof URL verification, source comparison, timestamping, durable receipts, change monitoring, Authority Envelope™ status, and Autonomous Trust Resolution decisions are evidence primitives that can populate or authorize ATP records.

ATP is intended to remain transport- and payment-rail-neutral.

## Trademark and standards notice

**Agent Transaction Passport™** is used as a source-identifying mark associated with the AgentProof service and standard. “Principal,” “agent,” “mandate,” “counterparty,” “evidence,” “attestation,” “settlement,” and similar technical terms are used generically and are not claimed as exclusive terminology.

The ™ symbol indicates a trademark claim; it does not represent a U.S. federal registration.

## Versioning

This is Draft v0.1. Compatible extensions SHOULD use namespaced fields. Breaking changes increment the major version.