Sandbox

Active

An AI agent for Sandbox. It works in a scene with a live connection to Sandbox. It asks before it writes, and every run is traced.

Run code in secure cloud sandboxes. Execute Python, JavaScript, shell commands, and more in isolated environments. Perfect for data analysis, code generation, and testing.

Infrastructure API Key 2 asset types 10 tools

What you can connect

Add these to your scene and AI gets access.

Account Accounts

Sandbox account for cloud code execution

Sandbox Sandboxes

A persistent code execution environment — auto-pauses when idle, resumes instantly

Tools

What the agent can do once this is connected.

Writes 10

Change data in the connected system.

sandbox_run Run a shell command in the sandbox.

sandbox_write Write content to a file.

sandbox_read Read file contents.

sandbox_list List directory contents.

sandbox_info Get sandbox status and info.

sandbox_get_url Get the public URL for a port running inside the sandbox.

sandbox_pause Pause the sandbox to save costs.

sandbox_destroy Permanently destroy the sandbox.

sandbox_create Create a new E2B sandbox.

sandbox_list_all List all E2B sandboxes on the account.

Connect Sandbox over MCP

Daslab is itself an MCP server.

Connect Sandbox in a scene, then pin that scene's URL in Claude Code, Cursor, or any MCP client. The agent there gets your Sandbox tools, and your approval gate and trace travel with them.

https://daslab.run/mcp/<your-world>/<your-scene>

One scene can hold Sandbox and everything else you connect, so a single endpoint carries them all. Set up a client.

Use cases

Execute generated code safely in sandboxes
Run data analysis scripts with pandas/numpy
Test code snippets before committing

Similar integrations

Others in Infrastructure.

SSH

Ready to try Sandbox with Daslab?

Get started with CLI