Credential Delegation Protocol for AI Agents in Multi-System Environments
Credential Delegation Protocol for AI Agents in Multi-System Environments is an agentic-web standard (endpoint) at step 3, Permit, governed by Individual Internet-Draft submitted to the IETF WIMSE Working Group mailing list; its own status is "Internet-Draft, Intended Status: Standards Track; expires 16 April 2027" (draft). It answers: How can a human user delegate scoped, attenuated, revocable credentials to AI agents across multiple service providers without exposing the underlying OAuth tokens to the agents? Adoption is unmeasured, last verified 2026-10-18.
Permit An individual Internet-Draft in the IETF WIMSE space that profiles OAuth 2.0 Token Exchange (RFC 8693), DPoP (RFC 9449), Rich Authorization Requests (RFC 9396) and CIBA for delegating scoped credentials to AI agents. It defines ephemeral-key agent identities, capability-shaped delegation tokens, credential wrapping, consent-gated delegation, cascading revocation and audit chains. It defines no new token formats or grant types and is positioned as a companion to draft-ietf-wimse-aims.
Discovery and probe
Delegation Server endpoints within a trust domain; no standard discovery path described in the available text
NONE parse: on absent: No externally observable site-level signal described; treat support as unknown
History
- 2026-10-16 — auto-drafted candidate from exa (https://datatracker.ietf.org/doc/html/draft-sweeney-wimse-credential-delegation-01)
Instances
No instances recorded yet.
See also
References
- Internet-Draft draft-sweeney-wimse-credential-delegation-01, published 13 October 2026, Intended Status Standards Track, workgroup WIMSE; profiles RFC 8693, RFC 9449, RFC 9396 and OpenID CIBA for credential delegation to AI agents. — https://datatracker.ietf.org/doc/html/draft-sweeney-wimse-credential-delegation-01 (2026-10-13) REPORTED
- Source for the draft and issue tracker are hosted on GitHub; the reference implementation is reported to pass twenty Asor interop vectors. — https://github.com/cred-ninja/protocol REPORTED