Cursor Rules MDC Generator — Auto-Generate from Docs

What it is

Cursor Rules MDC Generator is a Python tool that creates Cursor AI rule files (.mdc) for any programming library. It uses Exa semantic search to find documentation, best practices, and coding patterns for a specified library, then feeds that information through an LLM (Gemini, OpenAI, or Anthropic) to produce structured rule files that Cursor follows during code generation.

This tool targets developers who want Cursor to follow library-specific conventions automatically rather than repeating the same instructions in every conversation. Instead of manually writing .mdc files, you generate them from the library's actual documentation.

How it saves time or tokens

Manually writing a comprehensive .mdc rule file for a framework like FastAPI or Next.js takes hours of reading documentation and encoding conventions. This generator does it in minutes by crawling real documentation via Exa and distilling it into structured rules. The generated rules then save tokens in every future Cursor session by reducing the need to re-explain conventions.

How to use

1. Clone the repository and install dependencies: git clone https://github.com/sanjeed5/awesome-cursor-rules-mdc.git && cd awesome-cursor-rules-mdc && uv sync.
2. Add your API keys to a .env file (EXA_API_KEY plus your LLM provider key).
3. Generate rules for a library: uv run src/generate_mdc_files.py --library react.

Example

# Generate rules for FastAPI
uv run src/generate_mdc_files.py --library fastapi

# Generate rules for a category
uv run src/generate_mdc_files.py --tag backend

# Batch generation with concurrency
uv run src/generate_mdc_files.py --workers 4 --rate-limit 10

# Test mode (preview without writing)
uv run src/generate_mdc_files.py --library react --test

Output is a .mdc file you drop into .cursor/rules/ in your project.

Related on TokRepo

• AI Tools for Coding -- explore coding assistants and rule systems for AI-powered development
• Prompt Library -- browse reusable prompts and configuration templates

Common pitfalls

• Exa API has rate limits; batch generation with high concurrency may hit throttling. Use the --rate-limit flag to stay within bounds.
• Generated rules reflect documentation at the time of generation; re-run the generator when libraries release major version updates.
• The quality of generated rules depends on the LLM used; Claude and GPT-4 produce more nuanced rules than smaller models.