Ruby Agent

This is a coding agent implemented in Ruby, forked from the experiment accompanying this blog post: "Coding agent in 94 lines of Ruby".

Usage

Whatever method you use, first copy the .env.example file to .env and add your Anthropic API key. If you want to use a different provider, modify the run.rb file and set the key for the other provider. Check RubyLLM configuration documentation for details.

Model Context Protocol (MCP) Support

This agent now supports Model Context Protocol (MCP) which allows you to add external tools. To use MCP tools, configure the mcp.json file in the root directory:

{
  "mcpServers": {
    "serverName": {
      "command": "command-to-run-mcp-server",
      "args": ["arg1", "arg2"]
    }
  }
}

For example, to use the Playwright MCP server (included by default):

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@playwright/mcp@latest"]
    }
  }
}

You can also use an SSE-based MCP server by specifying a URL:

{
  "mcpServers": {
    "webMCP": {
      "url": "http://localhost:9292/mcp/sse"
    }
  }
}

To check if your MCP configuration is working, run:

ruby check_mcp.rb

With docker

If you have docker the usage is really simple. Just run the run_in_docker.sh script.

The directory from which you run the script will be mounted into the container as /workspace and will be the directory in which the coding agent will operate.

Without docker

If you're running it without docker you'll need Ruby and bundler installed.

Navigate to the root of the directory, run bundle install.

After that, call ruby /path/to/run.rb from the directory you want to operate on.