# 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](https://www.youtube.com/watch?v=qoaeYMrXJH0) on YouTube with a few slight mods.

This demo uses [Podman](https://podman.io/) 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
```