Meeting Notes 2026‐07 16 - openid/authzen GitHub Wiki
Meeting Notes - 2026-07-16
Attendees
- Alex Olivier
- Alex Babeanu
- Karl McGuinness
- Omri Gazitt
- Atul Tulshibagwale
- Vladi Berger
- John Jiang
- Edmund Jay
- Dave Hyland
- Nate Yocom
Agenda
- (Chairs) New draft proposal process
- (AlexO) Final points on COAZ + MCP Binding before Draft 2 (https://github.com/openid/authzen/pull/541)
- (AlexB) Obligations
- (Karl) ARAP/AROP Proposals
Notes
Draft Proposal Process
- Draft your proposal (Google Docs prefered) and contact co-chairs who will add to the OpenID WG AuthZEN Google Drive
- Share with WG via mailing list
- Co-chair's will make time on WG call agenda to present the proposal
- WG will give feedback
- If WG agrees to go forward, there will be a vote on the mailing list
- Should the vote passed, then the draft can be moved to
openid/authzenrepo (co-chairs can assist)
Questions:
- Atul: Slack vs mailing list?
- Karl: What is this process once it is in the WG?
- Omri: Notice a week in advance and have vote on call avoid mailing list
- Karl: When a WG draft gets adopted what are the minimal roles/responsibilties required for moving it forward as a distinct thread.
Wrapping up COAZ
- Subject trust anchor — Customer-service-bot case: agent acts for a user who isn't in the token. Current draft: subject.id MUST come from the token; the customer rides as context/resource attributes. Should we lossen to "at least one subject field from the token"?
- The agent idenity is going to be somewhere in the AuthZEN call.
- John: In OAuth, the subject should be the user and the agent would be the owner.
- Omri: Easier to say the subject.id MUST come from the token from a validation perspective. Could be the
actclaim (chain of subjects).subject.idSHOULD be thetoken.sub. - Atul: Lossen it to 'something' must come from the token.
- Karl: Most of the mapping logic is in the app layer before getting down to the MCP layer
- Atul: What we don't have is a standardised mapping of the token to the SARC call.
- Omri: Proposed default is the
subclaim in the token and have language which says if the subject has been overriden, the ID SHOULD come from a feild in the token. - Vladi - what about different/multiple tokens - do you have lots of mapping logic. CEL/tenaries.
- Karl - need a domain mapping layer to pick out/parse tokens. To address later in a further revision of the COAZ-MCP binding of the framework.
- Per-tool opt-out?
- Don't allow - remove from spec
Process point - what do number/version these drafts as?
- Both land as Draft 1 and deprecate the existing MCP. Also update the landing page.
Obligations
https://openid.github.io/authzen/authzen-obligations-profile-1_0.html
- Have a focused session week after next when Alex B back.
ARAP PR/Issues
- Search APIs provide no way to discover requestable resources https://github.com/openid/authzen/issues/555
- Add AuthZEN Access Request OAuth Profile (AROP) https://github.com/openid/authzen/pull/531
- scope evaluation_id-only denial binding to shared-state deployments https://github.com/openid/authzen/pull/532
- ARAP does too many things https://github.com/openid/authzen/issues/520
Karl: being doing work on the OAuth layer to support async which then can call ARAP.
Next week: ARAP and AROP week