MDC Rules Generator

Disclaimer: This project is not officially associated with or endorsed by Cursor. It is a community-driven initiative to enhance the Cursor experience.

This project generates Cursor MDC (Markdown Cursor) rule files from a structured JSON file containing library information. It uses Exa for semantic search and LLM (Gemini) for content generation.

Features

• Generates comprehensive MDC rule files for libraries
• Uses Exa for semantic web search to gather best practices
• Leverages LLM to create detailed, structured content
• Supports parallel processing for efficiency
• Tracks progress to allow resuming interrupted runs
• Smart retry system that focuses on failed libraries by default

Prerequisites

• Python 3.8+
• uv for dependency management
• API keys for:
  • Exa (for semantic search)
  • LLM provider (Gemini, OpenAI, or Anthropic)

Installation

1. Clone this repository:

   git clone https://github.com/sanjeed5/awesome-cursor-rules-mdc.git
   cd awesome-cursor-rules-mdc
   

2. Install dependencies using uv:

   uv sync
   

3. Set up environment variables: Create a .env file in the project root with your API keys (see .env.example):

   EXA_API_KEY=your_exa_api_key
   GEMINI_API_KEY=your_google_gemini_api_key  # For Gemini
   # Or use one of these depending on your LLM choice:
   # OPENAI_API_KEY=your_openai_api_key
   # ANTHROPIC_API_KEY=your_anthropic_api_key
   

Usage

Run the generator script with:

uv run src/generate_mdc_files.py

By default, the script will only process libraries that failed in previous runs.

Command-line Options

• --test: Run in test mode (process only one library)
• --tag TAG: Process only libraries with a specific tag
• --library LIBRARY: Process only a specific library
• --output OUTPUT_DIR: Specify output directory for MDC files
• --verbose: Enable verbose logging
• --workers N: Set number of parallel workers
• --rate-limit N: Set API rate limit calls per minute
• --regenerate-all: Process all libraries, including previously completed ones

Examples

# Process failed libraries (default behavior)
uv run src/generate_mdc_files.py

# Regenerate all libraries
uv run src/generate_mdc_files.py --regenerate-all

# Process only Python libraries
uv run src/generate_mdc_files.py --tag python

# Process a specific library
uv run src/generate_mdc_files.py --library react

Adding New Rules

Adding support for new libraries is simple:

1. Edit the rules.json file:

   • Add a new entry to the libraries array:

   {
     "name": "your-library-name",
     "tags": ["relevant-tag1", "relevant-tag2"]
   }
   

2. Generate the MDC files:

   • Run the generator script:

   uv run src/generate_mdc_files.py
   

   • The script automatically detects and processes new libraries

3. Contribute back:

   • Test your new rules with real projects
   • Consider raising a PR to contribute your additions back to the community

Configuration

The script uses a config.yaml file for configuration. You can modify this file to adjust:

• API rate limits
• Output directories
• LLM model selection
• Processing parameters

Project Structure

.
├── src/                  # Main source code directory
│   ├── generate_mdc_files.py  # Main generator script
│   ├── config.yaml       # Configuration file
│   ├── mdc-instructions.txt   # Instructions for MDC generation
│   ├── logs/             # Log files directory
│   └── exa_results/      # Directory for Exa search results
├── rules-mdc/            # Output directory for generated MDC files
├── rules.json            # Input file with library information
├── pyproject.toml        # Project dependencies and metadata
├── .env.example          # Example environment variables
└── LICENSE               # MIT License

License

MIT License