Sandboxes · Create a sandbox
Docs / Sandboxes

Create a sandbox

Create a sandbox from the console by picking a template and resources, then manage its lifetime from the Sandboxes page.

Create a sandbox from the console in a few clicks: pick a template, choose the resources it gets, and it's ready to use.

Create one

  1. 1
    Open Sandboxes
    From the dashboard sidebar, open the Sandboxes page.
  2. 2
    Click New Sandbox
    This opens the create form.
  3. 3
    Pick a template
    Choose a built-in template (Python, Node.js, Bash, LaTeX, AI Agent, Browser…) or one of your own custom templates as the starting image.
  4. 4
    Set resources and options
    Choose vCPUs and memory, a network mode and TTL, and optionally allowed domains, environment variables, an exposed port, and a persistent /workspace volume.
  5. 5
    Create
    Click Create. The sandbox boots immediately and opens to its console — no waiting on a separate provisioning step. A cost estimate (billed per hour from your credits) is shown before you launch.
ParameterTypeDescription
TemplaterequiredselectThe starting image — a built-in template or one of your own. See Templates.
vCPUs / Memory MBnumberCPU and memory allotted to the sandbox. Defaults to 1 vCPU / 512 MB.
Network Modeselectrestricted (default), full (open egress), proxied (routed through a proxy), or none (no network).
TTLselectHow long the sandbox lives before it is reclaimed — 15 min, 30 min, 1 hour, 2 hours, or 24 hours.
Allowed DomainstextComma-separated domains the sandbox may reach — pairs with a restricted network mode.
Environment VariablestextKEY=VALUE pairs, one per line, injected into the sandbox.
Expose Container PortnumberOptionally forward a host port to a container port, to reach a server running inside the sandbox.
Persistent StoragetoggleMount a /workspace volume that persists across sandbox restarts.
The New Sandbox dialog with template, resources, network mode, TTL, and storage options
The New Sandbox dialog — pick a template, then set resources, network mode, TTL, and storage.

What resources and lifetime actually mean

These two choices are about matching the sandbox to the job, not about picking the biggest option by default:

  • Resources — a small CPU/memory allotment is plenty for trying a snippet or poking at a small script. Reach for more when the job is heavier: processing a large dataset, running a memory-hungry library, or compiling something. Under-provisioning shows up as slow runs or out-of-memory errors, not a hard block, so it's easy to bump up and recreate if a job turns out bigger than expected.
  • Lifetime — a short lifetime is fine for a quick, one-off check you'll finish in the same sitting. Set a longer lifetime for something you're actively coming back to over a few days, so it isn't reclaimed between sessions. Either way, Extend is always available if you're not done yet, and a snapshot protects the state if you're not sure you'll get back to it in time.

Start from a snapshot instead

Instead of a template, you can start a new sandbox from an earlier snapshot — it boots with the same files and setup already in place, so you skip redoing any setup work.

Managing existing sandboxes

The Sandboxes page lists every sandbox you've created with its status. Open one to use it, or click Extend to keep it alive longer, or Delete to remove it immediately.

Tip
Deleting a sandbox removes its files right away. If you might want that state again, take a snapshot first.