Devin Cursorrules — Turn Cursor into an AI Engineer
A .cursorrules configuration that gives Cursor/Windsurf Devin-like capabilities: automated planning, self-correction, web browsing, and multi-agent collaboration. 6,000+ GitHub stars.
TO
TokRepo精选 · Community
Quick Use
Use it first, then decide how deep to go
This block should tell both the user and the agent what to copy, install, and apply first.
**Recommended (Cookiecutter):**
```bash
pip install cookiecutter
cookiecutter gh:grapeot/devin.cursorrules --checkout template
```
**Manual:**
1. Clone the repository:
```bash
git clone https://github.com/grapeot/devin.cursorrules.git
```
2. Copy `.cursorrules` to your project root (for Cursor)
3. Copy `.windsurfrules` + `scratchpad.md` (for Windsurf)
4. Copy `.github/copilot-instructions.md` (for GitHub Copilot)
5. Copy the `tools/` folder to your project
6. Install dependencies: `pip install -r requirements.txt`
---
Intro
Devin Cursorrules transforms Cursor, Windsurf, or GitHub Copilot into a Devin-like autonomous AI software engineer — without the $25/task pricing. With 6,000+ GitHub stars and MIT license, it adds automated task planning, self-correcting behavior, web browsing/search integration, screenshot verification via Playwright, and optional multi-agent collaboration (o1 for planning, Claude/GPT-4o for execution). Drop the config files into any project and your AI editor starts thinking before it acts.
Best for: developers who want Devin-level AI coding capabilities in Cursor or Windsurf for free. Works with: Cursor, Windsurf, GitHub Copilot. Setup time: under 3 minutes.
---
## Devin Cursorrules — How It Works
### Core Capabilities
#### 1. Automated Planning & Self-Evolution
The AI creates a structured plan before writing code, tracks progress in `scratchpad.md`, and updates its strategy as it learns from the codebase. Unlike vanilla Cursor, the agent thinks step-by-step and adapts its approach.
#### 2. Extended Tools
The `tools/` folder adds capabilities beyond basic code editing:
- **Web browsing** — Fetch and analyze web pages
- **Search engine integration** — Look up documentation and solutions
- **Screenshot verification** — Use Playwright to visually verify UI changes
- **LLM analysis** — Send data to external LLMs for deeper reasoning
#### 3. Multi-Agent Collaboration
Optional setup where:
- **o1/o1-pro** handles high-level planning and strategy
- **Claude/GPT-4o** executes the plan and writes code
- Agents communicate through shared scratchpad files
#### 4. Self-Correction
When you correct the AI ("no, use X instead"), it records the feedback and avoids the same mistake in future interactions within the session.
### File Structure
```
.cursorrules # Cursor AI configuration rules
.windsurfrules # Windsurf equivalent
scratchpad.md # Planning & progress tracking
.github/
copilot-instructions.md # GitHub Copilot config
tools/
browser.py # Web browsing capability
search.py # Search engine integration
screenshot.py # Playwright screenshot verification
requirements.txt # Python dependencies
```
### Supported Editors
| Editor | Config File | Status |
|--------|------------|--------|
| Cursor | `.cursorrules` | Full support |
| Windsurf | `.windsurfrules` + `scratchpad.md` | Full support |
| GitHub Copilot | `.github/copilot-instructions.md` | Basic support |
### FAQ
**Q: What is devin.cursorrules?**
A: A configuration that gives Cursor/Windsurf Devin-like AI capabilities — automated planning, self-correction, web browsing, and multi-agent collaboration — for free.
**Q: Is it free?**
A: Yes, MIT license. You only pay for the AI model tokens used by your editor (Cursor/Windsurf subscription).
**Q: How is this different from vanilla Cursor?**
A: Vanilla Cursor responds to individual prompts. With devin.cursorrules, the AI plans multi-step tasks, tracks progress, browses the web for context, verifies UI changes with screenshots, and learns from corrections.
---
🙏
Source & Thanks
> Created by [grapeot](https://github.com/grapeot). Licensed under MIT.
>
> [devin.cursorrules](https://github.com/grapeot/devin.cursorrules) — ⭐ 6,000+
Thank you to grapeot for making Devin-level AI coding accessible to every Cursor user.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.