Agent Credit Limits and Alerts - Copilot Studio

You need protection against rapid credit spikes and want to know whether limits act immediately. Microsoft documents monthly per-agent limits, approaching-limit notifications, and an optional hard stop. The agent is automatically turned off at the configured limit only when Stop usage is enabled; otherwise the limit status is monitoring, not enforcement.

Final consumption data is processed through the billing pipeline and can take several hours to appear. Microsoft publishes no evaluation interval or alert-delivery SLA, so neither is guaranteed real time.

For production, set a conservative limit below the true budget, enable Stop usage, assign a named admin owner, and review environment allocation and pay-as-you-go settings. Monitor estimated usage during testing, restrict high-frequency triggers and expensive actions, and keep a documented kill switch for disabling the agent. Validate the policy with a controlled non-production agent before relying on it for a high-volume workload.