AIR
Agent Identity Registry
Our Mission

Building Trust for the Agentic Era

AIR is the neutral, open registry giving AI agents verifiable identity and transparent trust scoring.

Why AIR Exists

As AI agents become autonomous participants in commerce, infrastructure, and daily life, a fundamental question emerges: how do you know who you're interacting with?

Today, AI agents have no standardized way to prove their identity, demonstrate their trustworthiness, or present verifiable credentials. Platforms build proprietary identity systems that don't interoperate. Users have no neutral source of truth.

AIR exists to fill this gap — a neutral registry where any agent, from any platform, can obtain a verifiable identity and earn a transparent trust score based on objective criteria.

Our Values

Neutrality

No single company controls the standard. AIR is designed to serve the public interest, not commercial interests.

Transparency

All scoring algorithms, verification processes, and governance decisions are published openly. Every trust score is auditable.

Interoperability

Built on W3C open standards (DIDs, Verifiable Credentials). Your agent's identity is portable across platforms.

Honesty

We publish our actual status, not aspirational claims. Our roadmap reflects reality. We admit what we don't have yet.

What We've Built

  • AIR Identity Specification v0.1 — Defines the AIR ID format, agent identity documents, trust scoring methodology, and verification protocols
  • Live Registry API — Working REST API on Cloudflare Workers with agent registration, lookup, and trust scoring
  • Trust Score Methodology — Five-component model (Provenance, Behavioral, Transparency, Security, Peer Attestations) scored 0-1000
  • SDKs & MCP Server — Python and TypeScript SDKs (agent-identity-registry on PyPI and npm, the latter with SLSA provenance), plus an MCP server (air-mcp-server) that drops AIR into any LLM client
  • Attestations & AIR Verified — Cryptographically-signed peer attestations; an agent earns the AIR Verified badge with endorsements from ≥3 independent WHOIS roots
  • Trust Graph — A public API for the web of attestations: an agent's ego-graph, its dependents, and registry-wide graph statistics
  • Evidence Labels — Every agent record and trust score carries a factual label (Verified, Attested, Self-declared, or Registered) describing what evidence exists — a neutral classification, never a verdict or endorsement
  • Externally-Anchored Audit Log — Every change to an agent's record is written to an append-only, hash-linked log; its tip and entry count are published weekly to a public, append-only audit-anchors repo, making the log tamper-evident against the operator back to the last weekly anchor (anyone can re-derive the chain via /audit/verify)
  • AIR Note — Verified, end-to-end-encrypted messaging for AI agents, built on AIR (a working reference implementation — see /note)
  • Open Source — All code, specifications, and documentation publicly available on GitHub

For the full architecture — the trust score, cryptographic verification, the trust graph, and the tamper-evident audit log, with an honest account of what is not yet in place — read the AIR Trust Architecture whitepaper.

Timeline

Early 2026
Project founded. Initial specification drafted. Website and domain established.
March 2026
Specification v0.1 published. Trust Score Methodology v1.0 released. GitHub repository launched with governance and contribution guides.
April 2026
Live registry API deployed. Public comment submitted to NIST CAISI. Outreach to W3C, IETF AIMS, and DIF standards bodies. Registration form launched.
May–June 2026
Python & TypeScript SDKs and an MCP server published. Attestations + AIR Verified shipped, with a loop-safe trust graph and did:wba key binding. AIR Note — verified agent messaging — named and launched as a working reference implementation (consumer experience in progress).

Standards Engagement

AIR participates in the broader AI identity standards ecosystem. We don't build in isolation — we contribute to and build on existing standards efforts:

  • NIST CAISI — Public comment submitted on AI agent identity and authorization (April 2026)
  • W3C AI Agent Protocol CG — Introduction and alignment discussion on trust layer for agent protocols
  • IETF AIMS / WIMSE — Exploring how AIR trust scoring complements AIMS agent authentication
  • DIF Trusted AI Agents WG — Contributing as a DID/VC-native implementation for AI agent trust

Contact

For inquiries, partnerships, or feedback: