Short answer: it resets monthly. - Prefect Cloud computes usage per workspace and the limit “resets monthly.” See: Managed infrastructure (compute) docs - Compute is measured as the time between the managed worker starting and tearing down your run, rounded up to the nearest minute (roughly your flow’s PENDING→COMPLETED window). Same doc link above. If you need an exact reset timestamp for your workspace or a temporary bump, please reach out to help@prefect.io. Also, if you’re considering upgrading from Hobby, I’m happy to connect you with our team to discuss Cloud/Horizon options and best practices for your use case—want me to make that intro?

Correct — managed/serverless compute is only consumed when a flow run executes on a Managed work pool (i.e., Prefect’s serverless infrastructure). - If a deployment targets a Managed work pool, any runs (scheduled or manual “Run” from the UI) will use managed compute credits. - If a deployment targets your own worker (Process, Docker, Kubernetes) or you run flows locally from your machine/CI, no managed compute is consumed. You’re just using Cloud for orchestration, logging, etc. Docs for reference: - Managed/serverless compute details: Managed infrastructure docs - Work pools overview (types and behavior): Work pools concepts If you want, I can also review your current pool/deployment setup to confirm which runs will (and won’t) count toward serverless compute.