lupin.Open console

Lupin / Sandboxes

Sandbox and container API for AI agents

Give an agent an isolated environment for code and files.

Build an agent workflow with sandboxes

Research, data processing and development tasks often need a working directory, tools and command output. Lupin’s hosted container API creates an isolated environment, exposes command results and lets the calling application stop or resume its runtime. It is distinct from the browser session API.

What the API supports

From setup to a verified outcome

  1. Check execution access

    Confirm that the hosted sandbox provider is configured and that the workspace has a sandbox runtime allowance and available concurrency.

  2. Run and verify

    Create or resume an environment, transfer the necessary files and run a bounded command. Inspect exit_code and the output; HTTP success alone does not mean the command succeeded.

  3. Control exposure and runtime

    Use application authentication for any published service and stop the environment when it is idle. Inspect usage rather than assuming a stopped browser also stops a sandbox.

Availability and limits

Check your workspace for its current configuration, usage and limits. Compare the plans before creating paid resources.

Start with the API reference

The following read endpoint is part of this capability. Authenticate with the relevant workspace key and inspect the documented response and permission requirements.

GET /v0/containers

OpenAPI specification · Sandboxes documentation · Agent registration

Common questions

Are browser sessions and sandboxes the same resource?

No. Browser sessions provide a browser environment. The container API provides an isolated place to execute commands and manage files.

Does a successful HTTP response mean my command worked?

No. Inspect the returned exit code, status, stdout and stderr to determine whether the command completed successfully.