Draft specifications — community review

Agent Trust Protocol Specifications A quantum-safe identity, trust, and privacy layer for AI agents.

ATP extends the W3C Decentralized Identifiers and Verifiable Credentials data models for autonomous agents. The four drafts below are ReSpec documents — click any to read the rendered specification. They cross-reference one another and are backed by a runnable reference implementation and conformance suite.

Identity

did:atp — Quantum-Safe DID Method

A DID method binding a classical Ed25519 key and a post-quantum ML-DSA-65 (FIPS 204) key in a single hybrid identifier, with deterministic pairwise derivation.

Draft · runnable proof + published test vectors
Trust

Agent Trust Scoring & Credentials

A deterministic, bounded, fail-closed trust model whose levels are expressible as signed, presentable W3C Verifiable Credentials (the ATP Trust Credential).

Early draft · seeking CG review
Privacy

Privacy-First Interaction

Pairwise (per-relationship) identifiers, Merkle-based selective disclosure with known-answer vectors, and experimental zero-knowledge range proofs.

Early draft · ZKP pending crypto review
Conformance

Conformance & Interoperability

Normative conformance items, audit-store integrity rules, and the vendor-neutral test-vector format that lets independent implementations verify interoperability.

Early draft