This repository provides a framework to imbue IDEs like Cursor and Windsurf, or GitHub Copilot, with advanced agentic AI capabilities, mimicking functionalities of premium tools like Devin at a significantly lower cost. It targets developers seeking to enhance their coding workflow with automated planning, self-evolution, and extended tool usage.

How It Works

The core mechanism involves customizing .cursorrules (or .windsurfrules and .github/copilot-instructions.md) files, coupled with Python scripts. This setup enables an AI agent to autonomously plan tasks, learn from feedback by updating its rules, and leverage tools such as web scraping (Playwright) and search engines (DuckDuckGo). An experimental multi-agent approach separates planning (o1) from execution (Claude/GPT-4o) for improved solution quality and iteration.

Quick Start & Requirements

• Install: pip install cookiecutter then cookiecutter gh:grapeot/devin.cursorrules --checkout template (Recommended) or manual file copying.
• Prerequisites: Python, Cursor/Windsurf IDE or GitHub Copilot. API keys for LLMs are optional but recommended. Playwright browsers are installed automatically on first use for screenshot features.
• Setup Time: Under a minute.
• Links: Step-by-step tutorial, Blog post

Highlighted Details

• Mimics Devin's automated planning and self-evolution.
• Integrates web browsing, search, and LLM-based analysis.
• Experimental multi-agent system for planning and execution.
• AI can learn from user corrections and update its .cursorrules.

Maintenance & Community

• No specific contributors, sponsorships, or roadmap details are highlighted in the README.

Licensing & Compatibility

• License: MIT
• Compatibility: Permissive for commercial use and integration with closed-source projects.

Limitations & Caveats

The multi-agent collaboration feature is explicitly marked as experimental. The effectiveness of self-evolution and tool usage is dependent on the underlying LLM's capabilities and the quality of the .cursorrules configuration.