RAT MCP Server (Retrieval Augmented Thinking)

A Model Context Protocol (MCP) server that implements RAT's two-stage reasoning process, combining DeepSeek's reasoning capabilities with various response models.

Features

• Two-Stage Processing:

  • Uses DeepSeek for detailed reasoning and analysis
  • Supports multiple models for final response generation
  • Maintains conversation context between interactions

• Supported Models:

  • DeepSeek Reasoner (for thinking process)
  • Claude 3.5 Sonnet (via Anthropic)
  • Any OpenRouter model (GPT-4, Gemini, etc.)

• Context Management:

  • Maintains conversation history
  • Includes previous Q&A in reasoning process
  • Supports context clearing when needed
  • Configurable context size limit

Installation

1. Clone the repository:

git clone https://github.com/newideas99/RAT-retrieval-augmented-thinking-MCP.git
cd rat-mcp-server

2. Install dependencies:

npm install

3. Create a .env file with your API keys and model configuration:

# Required: DeepSeek API key for reasoning stage
DEEPSEEK_API_KEY=your_deepseek_api_key_here

# Required: OpenRouter API key for non-Claude models
OPENROUTER_API_KEY=your_openrouter_api_key_here

# Optional: Anthropic API key for Claude model
ANTHROPIC_API_KEY=your_anthropic_api_key_here

# Optional: Model configuration
DEFAULT_MODEL=claude-3-5-sonnet-20241022  # or any OpenRouter model ID
OPENROUTER_MODEL=openai/gpt-4  # default OpenRouter model if not using Claude

4. Build the server:

npm run build

Usage with Cline

Add to your Cline MCP settings (usually in ~/.vscode/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json):

{
  "mcpServers": {
    "rat": {
      "command": "/path/to/node",
      "args": ["/path/to/rat-mcp-server/build/index.js"],
      "env": {
        "DEEPSEEK_API_KEY": "your_key_here",
        "OPENROUTER_API_KEY": "your_key_here",
        "ANTHROPIC_API_KEY": "your_key_here",
        "DEFAULT_MODEL": "claude-3-5-sonnet-20241022",
        "OPENROUTER_MODEL": "openai/gpt-4"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Tool Usage

The server provides a single tool generate_response with the following parameters:

{
  "prompt": string,           // Required: The question or prompt
  "showReasoning"?: boolean, // Optional: Show DeepSeek's reasoning process
  "clearContext"?: boolean   // Optional: Clear conversation history
}

Example usage in Cline:

use_mcp_tool({
  server_name: "rat",
  tool_name: "generate_response",
  arguments: {
    prompt: "What is Python?",
    showReasoning: true
  }
});

Development

For development with auto-rebuild:

npm run watch

License

MIT License - See LICENSE file for details.

Credits

Based on the RAT (Retrieval Augmented Thinking) concept by Skirano, which enhances AI responses through structured reasoning and knowledge retrieval.