MCP Sandbox

🚀 MCP Sandbox

MCP Sandbox is a Python-based sandbox environment tool that executes Python code in a securely isolated environment. It uses Docker container technology to ensure each task runs independently, preventing malicious code from damaging the host system.

🚀 Quick Start

MCP Sandbox is a Python-based sandbox environment tool designed to execute Python code in a securely isolated environment. It leverages Docker container technology to ensure that each task runs independently, preventing malicious code from causing damage to the host system.

📦 Installation

1. Clone the Repository Use the Git command to clone the MCP Sandbox project to your local machine:

   git clone https://github.com/your-username/mcp-sandbox.git
   

2. Create a Virtual Environment and Synchronize Dependencies After entering the project directory, install the UV package manager and synchronize the dependencies:

   uv create venv --name mcp-sandbox
   cd venv && uv sync
   

3. Start the Server Start the MCP Sandbox server:

   uv run main:app
   

   Then, you can connect to and use the MCP Sandbox tool by accessing http://localhost:8000/sse.

✨ Features

• Docker Isolation: Each sandbox environment runs in an independent Docker container, ensuring that tasks do not interfere with each other.
• Package Management: Supports the installation of third-party Python packages in the specified sandbox, such as numpy and pandas.
• File Generation and Access: All generated files (e.g., images, CSVs) are provided via HTTP links for direct access, facilitating display and embedding.

💻 Usage Examples

Workflow

1. Create or Select a Sandbox

   mcp_sandbox list_sandboxes
   

   If no sandboxes are available, create a new one:

   mcp_sandbox create_sandbox
   

2. Install Necessary Packages Install the required packages into the specified sandbox:

   mcp_sandbox install_package --sandbox_id <your-sandbox-id> --package numpy
   

3. Execute Python Code Run Python code in the sandbox:

   mcp_sandbox execute_python_code --sandbox_id <your-sandbox-id> --code "import pandas as pd; print(pd.__version__)"
   

4. View Results and Files The output, error messages, and generated file links after execution will be directly returned for further processing and display.

Example Code

The following is a complete example:

import pandas as pd

# Read data and save it as a CSV
data = {'A': [1, 2, 3], 'B': [4, 5, 6]}
df = pd.DataFrame(data)
df.to_csv('output.csv', index=False)

# Save an image using plt.savefig()
import matplotlib.pyplot as plt
plt.plot([1, 2, 3], [4, 5, 6])
plt.savefig('graph.png')

📚 Documentation

Configuration Example for Claude

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

Online Demo Configuration

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

⚠️ Important Note

Please adjust the serverUrl parameter according to your environment.