Policies And Approvals

Control-plane resources determine whether a requested action proceeds as submitted or requires an explicit governance outcome first. The public API exposes policy rules, approval cases, and action tokens as separate resources.

What Belongs Here

Policy rules declare conditions and a decision for an operation type at a scope. Create a rule with POST /policy-rules, edit its draft with PATCH /policy-rules/:id, then activate it. Update, activate, and retire require If-Match for the current version. Reads use policies.read, creation uses policies.create, and lifecycle changes use policies.manage; target scope checks also apply.

Approval cases hold explicit decisions when governed execution requires review. List or read cases, discover eligible approvers, then approve, reject, or cancel through the corresponding action endpoint. Decision capability alone does not establish eligibility: the service also enforces scope coverage and the case's separation-of-duties rules.

The operation exposes pending_approval and approval_ref so an integration can follow the business action while its governance case is resolved.

Action tokens authorize a specific command for a human caller. Issue one with POST /action-tokens and submit it in X-Action-Token with the matching protected command. Issuance and approval decisions can require MFA step-up; follow an insufficient_user_authentication challenge before retrying. Service principals are exempt from the human action-token requirement.

The model is built so governance is part of the runtime contract — an integration sees governance outcomes through the same operation and event streams it already consumes, not through a separate administrative side-channel. See the Governance And Events kickstart for request examples drawn from the published contract.