Cross Domain · security identity · IETF — individual submission, not yet working-group adopted

AI Agent Authentication and Authorization (draft-klrc-aiagent-auth)

An IETF Internet-Draft that proposes a composition — not a new protocol — for authenticating and authorizing AI agents, by treating an agent as a workload. It assigns each agent exactly one WIMSE identifier (which MAY be a SPIFFE ID), binds that identifier cryptographically with short-lived X.509 or JWT credentials issued off attestation evidence, authenticates calls at the transport layer with mTLS or at the application layer with WIMSE Proof Tokens or HTTP Message Signatures, and carries delegation through OAuth 2.0 — token exchange, transaction tokens, cross-domain identity chaining, and CIBA when a human has to approve. It names static API keys as an antipattern for agent identity.

IETF — individual submission, not yet working-group adoptedGoverning body
2026First published
draft-klrc-aiagent-auth-00 (2 March 2...Current version
No — informational framework, defines...Machine-readable schema · open
0providers in the catalog

Also known as draft-klrc-aiagent-auth, Agent Authentication and Authorization Framework, AI Agent Auth Framework, agent2agent-auth-framework.

Providers that carry AI Agent Authentication and Authorization 0

No provider in the catalog carries this standard as a tag yet. That is a statement about the catalog's tagging, not about adoption: a provider that implements AI Agent Authentication and Authorization and says so in its APIs.json profile appears here on the next build. Add or correct a listing.

Read from the API Evangelist standards directory, facts verified 2026-08-23. Provider links come from the catalog's own identity tags and are rebuilt nightly. If this page is wrong about a standard or a provider, say so.