A Ruby implementation of an AI Pair Programmer MCP (Model Context Protocol) server that provides AI-powered tools for code review, brainstorming, performance analysis, and security review.

This server provides 5 AI-powered tools:

1. pair - General collaboration and problem-solving
2. review - Comprehensive code review with actionable feedback
3. brainstorm - Creative ideation and solution exploration
4. review_performance - Performance analysis and optimization suggestions
5. review_security - Security-focused code review and vulnerability detection

• Ruby 3.0+
• An OpenRouter API key

The server uses inline bundler for automatic gem installation. Required gems:

• fast-mcp - Ruby MCP server framework
• ruby_llm - Unified AI model interface

1. Set your OpenRouter API key:

export OPENROUTER_API_KEY="your_api_key_here"

ruby ./server.rb

The server will:

• Automatically install missing gems on first run
• Start with STDIO transport for MCP clients
• Log to STDERR which should be saved by your MCP Host.

If the MCP fails to start when lunching Claude Code it's probably due to timeout. The bundler is installing dependecies. To fix that either set an envvar MCP_TIMEOUT=10000 (10s) or start the MCP server in the terminal first.

The server supports these AI models via OpenRouter:

• Gemini (default) - google/gemini-2.5-pro-preview
• O3 - openai/o3
• Grok - x-ai/grok-3-beta
• DeepSeek - deepseek/deepseek-r1-0528
• Opus - anthropic/claude-opus-4

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "AiPairProgrammer": {
      "command": "ruby",
      "args": ["/path/to/server.rb"]
    }
  }
}

Once connected to an MCP client:

Please review this Ruby code for best practices and potential issues:

def process_data(items)
  items.map { |item| item.upcase }
end

Can you analyze this code for performance bottlenecks?

def find_duplicates(array)
  duplicates = []
  array.each do |item|
    if array.count(item) > 1 && !duplicates.include?(item)
      duplicates << item
    end
  end
  duplicates
end

Please check this authentication code for security vulnerabilities:

def authenticate(username, password)
  user = User.find_by(username: username)
  if user && user.password == password
    session[:user_id] = user.id
    true
  else
    false
  end
end

I need ideas for improving user onboarding in my Ruby on Rails app. The current flow has a 60% drop-off rate.

I'm struggling with this algorithm problem. Can you help me think through it step by step?

The server is built with:

• FastMcp - Ruby MCP server framework with STDIO transport
• RubyLLM - Unified interface to AI models via OpenRouter
• Inline Bundler - Automatic gem installation for easy deployment

MIT License

AI Assistant MCP Server by Eduard https://github.com/eduardm/ai_pairs_with_ai

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test with ruby test_server.rb
5. Submit a pull request