Netmind Code Interpreter

You can run code in many languages securely in the cloud and access the results through a simple MCP client workflow. This NetMind Code Interpreter MCP Server lets agents execute multi-language code, manage files, and receive outputs like plots or data artifacts in a controlled, scalable environment.

How to use

To use the server with an MCP client, connect to the server’s MCP endpoint and issue the execute_code command through your client. You can run code in multiple languages, upload scripts and data files, provide input, and receive stdout, stderr, and any computed data as structured results. Outputs such as charts or images are automatically published as secure links.

How to install

Prerequisites: you need Python installed and access to install Python packages. You also need a cloud-enabled runtime to execute code securely and an API key for NetMind if you plan to use the provided online capabilities.

Install the UV tool (used to manage the code interpreter runtime): if you are on macOS, install via Homebrew or the install script; on Windows or Linux follow the provided installation methods.

Set your NetMind API token as an environment variable to authorize interactions with the NetMind backend.

Configure the MCP server entry in your Cursor/ Claude/ Windsurf setup by adding the following JSON snippet. This enables the code-interpreter MCP server to be started through the UV runtime and uses your API token for authentication.

{
  "mcpServers": {
    "code-interpreter": {
      "env": {
        "NETMIND_API_TOKEN": "XXXXXXXXXXXXXXXXXXXX"
      },
      "command": "uvx",
      "args": [
        "netmind-code-interpreter-mcp"
      ]
    }
  }
}

Additional setup you will perform

Start the server locally for development and testing using the Python module as shown in the example: you run the server with Python and then connect your MCP client to the given endpoint.

Test that the client can communicate with the server by running the provided test client. This ensures your environment can perform code execution, return outputs, and handle errors gracefully.

MCP client integration and endpoints

The MCP client should connect to the HTTP endpoint at the Server-Sent Events URL. The client uses the execute_code tool to send code files, language, and optional stdin and arguments.

Configuration and security notes

Environment variables shown here must be present in your runtime environment. Treat the NETMIND_API_TOKEN as a secret and keep it secure. Each execution runs in an isolated container with resource limits for safety and stability.

Troubleshooting

If you cannot connect from your MCP client, verify that the server is reachable at http://localhost:8000/sse, and ensure the API token is correctly set in your environment. Check that the uv runtime and the MCP server bundle are installed as shown in the installation steps.

Notes

You can upload scripts, CSVs, and other text-based files directly to the runtime. Your code can read, write, and process these files in a sandboxed filesystem for security and data integrity.

Available tools

execute_code

Execute code in a specified programming language inside a secure cloud execution environment. You provide the language, a set of files with names and contents, optional standard input, and optional command line arguments. The tool returns stdout, stderr, and any data outputs on success, or an error description on failure.