Skip to main content
Pause and resume a sandbox to keep its memory and files without running compute. Resume it to continue from retained state.

​
Pause and resume

Wait for the state transition before sending work. A pause keeps a checkpoint; it does not delete the machine. Restores keep the same CPU and memory bounds. Persistent machines pause after their idle timeout and wake on funded access through exec, files, desktop, or application HTTP. Temporary machines still have a deletion timeout; pausing does not turn them into indefinite retained workspaces.

​
Storage and funding

Paused machines pay for retained storage only. When a persistent machine cannot fund renewal, inspect its credit grace deadline and warnings. Top up before that deadline to keep it resumable. After the deadline the machine is deleted. Use prepaid credit for the funding model and GET /v1/pricing for current rates.

​
Check recovery state

A failed capture can preserve the source machine. Read preservation.message, operationID, recovered, and nextAction from machine detail. If recovered is true and nextAction is resume_or_delete_source, choose resume or delete according to your task. If recovery is still pending, inspect status before another capture attempt. billing: platform identifies a source whose tenant billing is paused for the failed operation. Lifecycle changes retire old desktop links and human control sessions. Open a new watch link after resuming.

​
Delete when finished

Deletion retires the machine. Save a reusable snapshot or download required files first.