Scale to zero is the deployment default
Agent tools should avoid always-on cloud cost unless the workload requires it and an estimate is approved.
Scale to zero is the deployment default
Production readiness includes cost behavior. For intermittent agent tools, a runtime that can scale to zero is usually the appropriate starting point.
Approval before persistence
Any always-on service should identify the reason it must remain warm, the expected request pattern, the minimum instance count, and a monthly estimate. The estimate is reviewed before the configuration changes.
Secrets and access
Runtime credentials belong in a managed secret store. Services receive only the secrets and roles required for their bounded task, and public access is disabled unless the product surface requires it.
Review status
This first-party operating note is suitable for template demonstration. Add external claims only through the capture, adversarial review, refinement, research, and approval workflow.