How Actera works

Actera evaluates intent before execution:

1. Your trusted integration identifies the agent and intended action.
2. It supplies an immutable context containing the facts needed by policy.
3. Actera selects published policies assigned to that agent and applicable to the action.
4. Every applicable rule is evaluated.
5. Actera resolves all contributing effects with BLOCK taking precedence over REQUIRE_APPROVAL, then ALLOW.
6. The integration enforces the result and, where supported, reports what happened afterwards.

If policies apply but none of their rules match, Actera blocks the action. If no policy applies, the agent’s configured fallback determines the result.

Approval and execution

An approval authorizes an action. It does not prove the external action completed. A trusted host must resume or dispatch the intended call after approval, preserve the action handle, and report a completed, failed, or uncertain outcome.

Testing without side effects

Policy tests use hypothetical context and return the same policy outcome and rule explanation as enforcement. They do not create authoritative decisions, approvals, usage reservations, or external actions. A test also cannot prove that a live usage reservation will still be available when concurrent actions arrive.