Introduction

ATAP (Agent Trust and Authority Protocol) is an open protocol for verifiable multi-party authorization in AI agent ecosystems. It defines how agents, machines, humans, and organizations can produce cryptographic proof of consent — portable, offline-verifiable, and built entirely on established standards.

Who is ATAP for?

ATAP is designed for protocol implementers: teams building agent platforms, identity infrastructure, payment systems, or any service where AI agents act on behalf of humans and need provable authorization.

If you're building:

• An agent framework that needs human-in-the-loop approvals
• A merchant system that accepts agent-initiated actions
• An identity provider that issues credentials to agents and humans
• A wallet or mobile app that renders approval requests

...then ATAP gives you a standard way to handle multi-party consent.

What ATAP provides

One novel contribution: multi-signature approvals where each party signs independently, producing a self-contained document that anyone can verify offline.

Everything else is composition of existing standards:

Core concepts

1. Entities — Every participant (agent, machine, human, organization) gets a did:web DID. See Entities & DIDs.

2. Approvals — The core document: a multi-signature authorization proof with 2 or 3 independent signatures. See Approvals.

3. Credentials — W3C Verifiable Credentials expressing verified properties (email, personhood, identity). Credentials determine trust level. See Credentials.

4. DIDComm — All entity-to-entity communication uses DIDComm v2.1 for encrypted messaging. See DIDComm.

5. Templates — Adaptive Cards define how approvals are rendered on devices. See Templates.

Next steps

• Quickstart — Register an entity on the sandbox and send your first approval.
• Protocol Specification — The full ATAP v1.0-rc1 spec.
• API Reference — All REST endpoints for the reference implementation.