Python Sandbox

🚀 MCP Sandbox

MCP Sandbox is a Docker - based secure sandbox environment designed for running and managing Python code, providing isolation and convenient package and file management.

🚀 Quick Start

You can try it out on mcp sandbox.

✨ Features

• Docker Isolation: Each executed task runs in an independent Docker container, ensuring environmental security.
• Package Management: Supports the installation and management of common Python packages (e.g., numpy, pandas, etc.).
• File Handling: Can generate and store various file types and provides direct access via HTTP links.

📦 Installation

1. Clone the repository:

   git clone https://github.com/yourusername/mcp-sandbox.git
   

2. Navigate to the project directory:

   cd mcp-sandbox
   

3. Start the service:

   npm run dev
   

💻 Usage Examples

Available Tools

The following are the main available tools and their functional descriptions:

• list_sandboxes(): Lists all existing sandbox environments.
• create_sandbox(): Creates a new sandbox environment and returns a new sandbox_id.
• delete_sandbox(sandbox_id): Deletes the specified sandbox environment.
• install_package(sandbox_id, package_name): Installs the specified Python package in the specified sandbox.
• check_install_status(sandbox_id, package_name): Checks the installation status of the specified package.
• execute_code(sandbox_id, code): Executes the provided Python code in the specified sandbox and returns the output, errors, and links to the generated files.

Example Workflow

1. Use list_sandboxes() to check the available sandbox environments. If no sandboxes are available, use create_sandbox() to create a new one and obtain the sandbox_id.
2. Use install_package(sandbox_id, package_name) to install the required packages (e.g., numpy, pandas, etc.).
3. Use check_install_status(sandbox_id, package_name) to verify the installation status of the package. If the installation is in progress, check again later.
4. Use execute_code(sandbox_id, code) to execute the code and get the output, errors, and links to the generated files.

📚 Documentation

MCP Example Configuration

The following is an example configuration for Claude:

{
  "mcpServers": {
    "mcpSandbox": {
      "command": "npx",
      "args": ["-y", "supergateway", "--sse",  "http://localhost:8000/sse"]
    }
  }
}

MCP Example Configuration for Online Demo

{
  "mcpServers": {
    "mcpSandbox": {
      "command": "npx",
      "args": ["-y", "supergateway", "--sse",  "http://115.190.87.78/sse?api_key=<API_KEY>"]
    }
  }
}

You may need to adjust the serverUrl according to your environment.