Code Sandbox MCP 🐳

What is Code Sandbox MCP?

Code Sandbox MCP is a secure sandbox server that executes code in isolated Docker containers. It provides AI applications with a safe environment for running Python, Go, Node.js, and TypeScript code while maintaining security through containerization.

How to use Code Sandbox MCP?

Install via the provided curl (Linux/Mac) or PowerShell (Windows) script, which checks for Docker, downloads the binary, and configures Claude Desktop automatically. Alternatively, manually download the binary from the releases page and add it to your PATH. Then use the run_code tool for snippets or run_project for full directory execution.

Key features of Code Sandbox MCP

• Multi-language support: Python, Go, Node.js, TypeScript.
• Automatic dependency detection and installation (pip, npm, go get).
• Real-time output and log streaming from containers.
• Background mode for long-running services.
• Volume mounting for project execution.
• Resource-limited, clean container cleanup after execution.

Use cases of Code Sandbox MCP

• Run untrusted code snippets safely without host system exposure.
• Execute full project directories with dependency management.
• Provide a secure code execution backend for AI assistants.
• Automatically test or debug code across multiple languages.

FAQ from Code Sandbox MCP

What languages are supported?

Python, Go, Node.js (including TypeScript/JSX via built-in Node.js support) are supported.

How does dependency management work?

Dependencies are automatically detected from import/require statements and installed via pip, npm, or go get. For projects, files like requirements.txt, go.mod, or package.json are used.

What are the prerequisites for using Code Sandbox MCP?

Docker must be installed and running on the host system. The installer verifies Docker availability.

How do I configure Code Sandbox MCP for Claude Desktop?

The automated installer adds the configuration to Claude’s config file. Manual configuration paths are provided for Linux (/.config/Claude/), macOS (/Library/Application Support/Claude/), and Windows (%APPDATA%\Claude\claude_desktop_config.json).

What security measures are in place?

Code runs in isolated Docker containers with resource limits, separate stdout/stderr streams, read-only project file mounts, and automatic container cleanup after execution.