Supercharge your Cursor AI with codebase-specific intelligence. Rulefy transforms your GitHub repositories into custom rules that teach Cursor AI how your project works, resulting in more accurate code suggestions, better refactoring, and contextually aware assistance.
- ✅ Instant Expertise: Your AI assistant immediately understands your project conventions and architecture
- ✅ Better Suggestions: Get code completions that match your project's style and patterns
- ✅ Fewer Hallucinations: Rules constrain the AI to work within your codebase's patterns and dependencies
- ✅ Zero Learning Curve: One command to analyze your repo and generate optimized rules
- 🚀 Analyze GitHub repositories or local codebases with a single command
- 🧩 Intelligently extract project structure, conventions, and patterns
- 📚 Handles large codebases that exceed the token limit for LLMs
- 🤖 Generate tailored Cursor rules using Claude AI
- 📝 Create production-ready .rules.mdc files for immediate use
- 🔧 Customize analysis with flexible configuration options
# Install globally
npm install -g rulefy
# Or use with npx (no installation required)
npx rulefy
- Node.js 18 or higher
- ANTHROPIC_API_KEY environment variable set with your Anthropic API key
export ANTHROPIC_API_KEY='your-anthropic-api-key'
The basic command structure is:
rulefy <repo-url-or-path>
Examples:
# Using GitHub URL
rulefy https://github.com/fastapi/fastapi
# Using local repository path
rulefy ./my-local-project
# Rulefy with specific description as an option
rulefy --description "guidelines for extending the component using the base interface"
# Rulefy the current directory with a description
cd some/package
rulefy --description "guidelines for extending the component using the base interface"
# Specify rule type
rulefy --rule-type "agent" --description "coding standards for React components"
This will:
- Fetch the repository content using repomix
- Read the cursor rules guidelines
- Generate cursor rules using Claude Sonnet 3.5
- Save the output to a
<repo-name>.rules.mdc
file
After generating your rules file, you'll need to install it in Cursor:
- Open Cursor editor
- Go to Settings > AI > Rules
- Click "Add Rules File" and select your generated
.rules.mdc
file - Restart Cursor to apply the new rules
For more detailed instructions, see the official Cursor documentation.
Options:
--provider <provider> LLM provider to use (default: "claude-sonnet-3.5-latest")
-o, --output <file> Output file name (defaults to <repo-name>.rules.mdc)
--guidelines <file> Path to cursor rules guidelines file (default: "./cursorrules-guidelines.md")
--include <patterns> Include patterns for files (glob pattern, e.g., "**/*.py")
--description <text> Description of what should be rulefied
--rule-type <type> Type of rule to generate (auto, manual, agent, always)
-h, --help display help for command
- Repository Conversion: Uses repomix to convert GitHub repositories into a textual representation
- Guidelines Integration: Reads cursor rules guidelines from the specified file
- LLM Generation: Passes the repository content and guidelines to LLM with progressive summarization
- Output: Saves the generated cursor rules to a markdown file
The generated .rules.mdc
file will contain structured guidelines for Cursor AI when working with your codebase:
# .cursorrules for my-project
## Project Overview
This project is a TypeScript application that...
## Coding Standards
- Follow TypeScript best practices with strict typing
- Use async/await for asynchronous operations
...
## File Structure Guidelines
- Place core logic in the `src/` directory
...
Contributions are welcome! Please feel free to submit a Pull Request.
This project is inspired by and builds upon the work of:
- repomix - A tool for converting repositories into textual representations
- awesome-cursorrules - A curated list of cursor rules for different projects and languages
We're grateful to these projects for their contributions to the developer tooling ecosystem.