Sample MCP Server in Go.

This MCP server runs a Python environment when called that can be used to execute the code generated by an LLM. It can be useful for scraping web content.

This project is based on this demo on YouTube with a few slight mods.

This demo uses Podman as the container engine instead of Docker.

Build the project

This project requires Go 1.23+. To build locally:

$ go mod tidy && go install

Testing with a lab host

For testing use mcphost: https://github.com/mark3labs/mcphost

Configure the MCP server

Edit the ~/.mcp.json file adding the following content:

{
  "mcpServers": {
      "python-repl": {
          "command": "python-mcp-server"
      }
  }
}

This configuration tells the mcphost to run the python-mcp-server command aliased as python-repl.

Running mcphost

This example uses Claude-3.5 Sonnet as the defaul LLM. This allows testing without a local GPU.

Export the Anthropic key:

$ export ANTHROPIC_API_KEY='your-api-key'

Execute the host

$ mcphost

It should prompt something like this:

$ mcphost                                                                                                                                        
2025/04/28 21:44:32 INFO Model loaded provider=anthropic model=claude-3-5-sonnet-latest
2025/04/28 21:44:32 INFO Initializing server... name=python-repl
2025/04/28 21:44:32 INFO Server connected name=python-repl
2025/04/28 21:44:32 INFO Tools loaded server=python-repl count=1
  Enter your prompt (Type /help for commands, Ctrl+C to quit)

From now on it is possible to prompt requests that imply a web scraping to see the MCP server in action.

  Enter your prompt (Type /help for commands, Ctrl+C to quit)
  Show me the most starred repositories on github

  Assistant:                                                                                                                                    

  I'll create a script to fetch the most starred repositories from GitHub using their API:                                                                                  

2025/04/28 22:14:16 INFO 🔧 Using tool name=python-repl__execute-python
2025/04/28 22:14:16 INFO Usage statistics input_tokens=1665 output_tokens=309 total_tokens=1974