Skip to main content
Use the operation ID returned by the mutation. queued or running means work remains; succeeded, failed and cancelled are terminal. Project status supplies actual environments, deployment heads and URLs. A queued operation with blocking_operation_id waits for that operation of the same project to end; inspect it instead of submitting again.

​
Observe the next step

During a deployment, progress.phase distinguishes building, publishing and waiting for mail. Follow progress.next_poll_after_seconds and the returned suggested action. Read sender or domain status when that dependency is pending. CLI --wait can reach its own timeout without cancelling the server operation. Read the same operation afterward. Never submit another deployment just to get a status update. For a live event stream:
MCP operation_logs returns a bounded snapshot. Use operation_get for current completion rather than assuming the snapshot waits until done.

​
Verify success

An operation marked succeeded must still produce the requested outcome. For up to thirty seconds after success, some requests can still reach the previous version or address; recheck before calling a change missing. For protected Dev, open the Owner’s Dev share link; for public Dev, open the clean URL. Then test the real application, or read back the changed domain, budget or compute settings. A data-change operation requires the named environment to be redeployed; follow its assignment plan. Errors and recovery · Limits.