Usage limits

Actera can enforce policies based on cumulative AI agent activity, not only individual actions. Usage limits apply rolling thresholds, historical counters, and other stateful constraints across multiple actions and workflows, such as a maximum refund total over 24 hours.

Assigned policies refer to named historical facts rather than querying external systems during rule evaluation. The trusted integration supplies any additional immutable context required by the rule.

For real enforcement, a matching action may reserve capacity atomically. Concurrent requests then observe the reservation instead of all reading the same stale total. A blocked, rejected, or failed action can release its reservation according to the configured lifecycle; a completed action commits it.

Policy tests are side-effect free. They can evaluate a supplied historical total, but they do not reserve capacity and cannot promise that capacity will remain available for a later live request. Review How Actera works for the full evaluation path and Decisions and approvals for reservation outcomes that require human review.