The request popped into the system at 14:03. It was a deploy-to-production action—dangerous if unchecked. Within seconds, Open Policy Agent (OPA) evaluated the request against a Just-In-Time (JIT) approval policy. The engineer’s credentials were valid, but OPA paused the action, waiting for a human reviewer to approve it in real time.
Open Policy Agent (OPA) Just-In-Time Action Approval is a method to enforce critical control over high-risk operations. Instead of blanket permissions, OPA policies define precise conditions under which an action can proceed. When a request matches those conditions, the system triggers a JIT workflow: ask for approval, confirm the context, then execute.
This pattern prevents privilege creep and stale permissions. Long-lived credentials are a liability. By shifting to JIT action approval with OPA, permission is granted only when needed and only for the task at hand. The policy is the single source of truth, evaluated the moment the request is made.
OPA runs as a lightweight decision engine. You define policies in Rego. The input data includes who is making the request, what the action is, when it is requested, and any environment context. The decision result returns "allow": true or "allow": false. For JIT action approval, the policy might return a "require_approval": true flag, triggering an external service to pause and request sign-off.