Sandboxes¶

Run superqode sandbox doctor before selecting a backend. Local execution is the default path; cloud providers are explicit integrations.

Core and popular backends¶

Sandbox	Install	Authentication	First run
Local OS	Included; uses macOS Seatbelt or Linux Bubblewrap when available	None	`superqode sandbox doctor local-os`
Docker	Install the Docker CLI and daemon	None	`superqode sandbox run docker --image python:3.12 -- pytest -q`
Podman	Install Podman	None	`superqode sandbox doctor podman`
Apple Container	Install the Apple `container` CLI; status support is experimental	None	`superqode sandbox doctor apple-container`
E2B	`uv tool install "superqode[sandbox-e2b]"`	`E2B_API_KEY`	`superqode sandbox run e2b -- "pytest -q"`
Daytona	`uv tool install "superqode[sandbox-daytona]"`	`DAYTONA_API_KEY`	`superqode sandbox doctor daytona`
Modal	`uv tool install "superqode[sandbox-modal]"`	Modal account configuration	`superqode sandbox doctor modal`
Vercel Sandbox	Install the Vercel Sandbox CLI	`VERCEL_OIDC_TOKEN` or `VERCEL_TOKEN`	`superqode sandbox doctor vercel`

Install the three bundled Python cloud providers together with:

uv tool install "superqode[sandbox-cloud]"

Experimental backends¶

Sandbox	Separate installation	Authentication	Verification
Runloop	`uv pip install runloop_api_client langchain-runloop`	`RUNLOOP_API_KEY`	`superqode sandbox doctor runloop`
Amazon Bedrock AgentCore	`uv pip install bedrock-agentcore langchain-agentcore-codeinterpreter`	AWS credentials	`superqode sandbox doctor agentcore`
LangSmith	`uv pip install langsmith deepagents`	`LANGSMITH_API_KEY`	`superqode sandbox doctor langsmith`

The OpenAI Agents runtime also recognizes optional agents-e2b, agents-daytona, agents-modal, agents-vercel, agents-runloop, agents-blaxel, and agents-cloudflare sandbox client packages. Install those directly into the active environment when using the OpenAI Agents sandbox surface.

See Sandbox Commands and Safety and Permissions.