Skip to main content
A snapshot captures a sandbox’s filesystem so you can restore it later. Do the expensive setup once (install dependencies, clone a repository, warm a cache), snapshot it, and start every later sandbox from that point instead of repeating the work.

Create a snapshot

Restore

Restoring produces a new sandbox with its own id. The snapshot is not consumed, so one snapshot can fan out into as many sandboxes as you need, which is what makes this useful for running a test matrix in parallel.
Fanning one snapshot out across several sandboxes:
Restored sandboxes count against the same 100-concurrent-sandbox limit as any other, so destroy them when the work finishes.

List and delete

Snapshot storage is not metered today, so keeping one costs nothing. Delete the ones you no longer need rather than relying on that. A restore forks a new sandbox, which counts against your concurrency limit and is billed for the CPU and memory it reserves.