AI-powered multi-repository collaboration system that works seamlessly with Claude Code, Cursor, GitHub Copilot, and more!
한국어 문서 | English
Working with AI on real projects? You face these issues:
1. Context is lost every session 😫
- New AI session = Start from scratch
- Explain the same architecture again and again
- "What's the API endpoint?" "How does auth work?" - Every. Single. Time.
2. Multi-repo chaos 🤯
my-saas-project/ ├── api-server/ (backend) ├── web-client/ (frontend) └── mobile-app/ (mobile) - AI only sees one repo at a time
- Missing context from other repos → Fragmented code
- "Add login" needs backend API + frontend UI, but AI doesn't know both
3. AI makes dangerous assumptions
- "I'll set the timeout to 30 seconds" - Wait, should be 5!
- "Using /api/v1/..." - Wrong endpoint!
- Guesses business logic, security settings, pricing rules
Result: You spend more time explaining and fixing than actual coding.
CodeSyncer gives AI the full picture through:
- 📝 Comments in code - All decisions and context live where they belong
- 🗂️ Master document - Cross-repo navigation and rules
- 📋 Per-repo docs - Each repo's specific guidelines
- 🎯 Keyword system - Auto-pause for critical decisions (payment, auth, etc.)
Result: AI codes with high accuracy even in complex multi-repo projects. 🎯
CodeSyncer provides the framework and rules for AI coding assistants (like Claude Code) to set up an intelligent collaboration system across your multi-repository workspace.
How it works:
- You install CodeSyncer CLI
- You launch your AI assistant (Claude Code, Cursor, etc.)
- You run
codesyncer init - AI analyzes your projects and generates documentation following CodeSyncer's structure
CodeSyncer defines WHERE and HOW documentation should be created. Your AI assistant fills in the WHAT by analyzing your actual code.
- 🤖 AI-Agnostic: Works with Claude Code, Cursor, GitHub Copilot, and more
- 📁 Single & Multi-Repository Support: Works with individual repos or entire workspaces
- 📦 Monorepo Support: Auto-detects Turborepo, pnpm, Nx, Lerna, Yarn/npm workspaces
- 🏷️ Comment Tag System:
@codesyncer-*tags to record decisions and inferences - 🤝 Discussion Auto-Pause: Automatically stops for critical decisions (payment, security, etc.)
- 🌐 Multi-Language: Full Korean and English support
- ⚡ Quick Setup: One-command installation for your entire workspace
CodeSyncer requires an AI coding assistant to be active.
Currently supported:
- ✅ Claude Code (Recommended)
- 🚧 Cursor (Coming soon)
- 🚧 GitHub Copilot (Coming soon)
- 🚧 Continue.dev (Coming soon)
Important: Make sure your AI coding assistant is running and active before using CodeSyncer. The AI will analyze your projects and help generate accurate documentation.
npm install -g codesyncercodesyncer --versionnpm view codesyncer versionnpm install -g codesyncer@latestWhen you update CodeSyncer to a new version, run the update command to sync your project with the latest templates and features:
cd /path/to/your/multi-repo-workspace codesyncer updateWhat happens:
- ✅ Scans for new repositories added to your workspace
- ✅ Detects missing files from new versions (e.g., root CLAUDE.md in v2.1.2+)
- ✅ Auto-detects your language settings (English/Korean)
- ✅ Prompts before creating any new files
- ✅ Preserves your existing customizations
Example output:
🔄 CodeSyncer - Update System ✓ Scan complete ⚠️ Missing root CLAUDE.md (new in v2.1.2) This file allows Claude to automatically load context at session start. ? Create root CLAUDE.md? (Y/n) Y ✓ Root CLAUDE.md created! 💡 Claude will now automatically load context at session start! 🤖 Next Steps (Tell your AI): ──────────────────────────────────────────────────────────── Option 1) Start a new session Claude will automatically find and read root CLAUDE.md Option 2) Apply immediately in current session "Read CLAUDE.md" ──────────────────────────────────────────────────────────── ✅ Update complete! After running codesyncer update:
Choose one of these options to apply changes:
Option 1: Start a new AI session (Recommended)
- Close your current AI assistant
- Open a new session
- Claude automatically finds and reads root CLAUDE.md
Option 2: Apply in current session
- Tell your AI: "Read CLAUDE.md"
- AI loads the updated context immediately
npm install -g codesyncerOpen your AI coding assistant:
- Claude Code (Recommended)
- Cursor
- GitHub Copilot
- Or any other AI coding tool
Make sure it's active and running.
cd /path/to/your/projectCodeSyncer works with single repositories, multi-repo workspaces, and monorepos:
Single Repository (auto-detected):
my-project/ ├── package.json ├── src/ └── ... Multi-Repository Workspace:
workspace/ ├── backend/ ├── frontend/ └── mobile/ Monorepo (auto-detected via Turborepo, pnpm, Nx, Lerna, npm/yarn workspaces):
monorepo/ ├── package.json # workspaces: ["packages/*", "apps/*"] ├── turbo.json # or pnpm-workspace.yaml, nx.json, lerna.json ├── packages/ │ ├── shared/ │ └── ui/ └── apps/ ├── web/ └── api/ codesyncer initYou'll be asked:
- Language preference (Korean/English)
- Project name
- GitHub username
What happens:
| Mode | Detection | Output |
|---|---|---|
| Single Repo | Current folder has package.json, .git, etc. | Creates .claude/SETUP_GUIDE.md |
| Monorepo | Has turbo.json, pnpm-workspace.yaml, nx.json, lerna.json, or workspaces in package.json | Creates .codesyncer/SETUP_GUIDE.md with package-aware setup |
| Multi-Repo | Subfolders contain separate repositories | Creates .codesyncer/SETUP_GUIDE.md |
That's all CodeSyncer does! It provides the framework and rules. Now your AI takes over.
🎯 Don't skip this step! This is where the magic happens.
Launch Claude Code (or your preferred AI assistant) and say:
For Single Repository:
"Read .claude/SETUP_GUIDE.md and follow the instructions to set up" For Multi-Repository Workspace:
"Read .codesyncer/SETUP_GUIDE.md and follow the instructions to set up" 1️⃣ AI Analyzes Your Code
- Reads actual files in each repository
- Detects tech stack, patterns, and structure
- Understands your project architecture
2️⃣ AI Asks Critical Questions (Never assumes!)
- ❓ "What are your API endpoints?"
- ❓ "What's your pricing and business logic?"
- ❓ "Which authentication method do you use?"
- ❓ "What's your database schema?"
- ❓ "Which external services do you integrate?"
3️⃣ AI Generates Complete Documentation
.codesyncer/MASTER_CODESYNCER.md→ Multi-repo navigation<repo>/.claude/CLAUDE.md→ Coding rules<repo>/.claude/ARCHITECTURE.md→ Project structure<repo>/.claude/DECISIONS.md→ Decision log<repo>/.claude/COMMENT_GUIDE.md→ Comment tag guide
💡 Why this works: AI analyzes YOUR actual code and asks YOU questions. The generated docs are tailored to your specific project, not generic templates.
Once setup is complete, just tell your AI:
"Read CLAUDE.md" Your AI assistant will then:
- Follow your project's coding rules
- Use the correct tech stack patterns
- Ask before making critical decisions
- Record all decisions with
@codesyncer-*tags
codesyncer initcodesyncer updatecodesyncer add-repoCodeSyncer uses a structured comment tag system to permanently record all AI inferences and decisions in your code.
| Tag | Purpose | Example |
|---|---|---|
@codesyncer-rule | Special rules for non-standard implementations | // @codesyncer-rule: Use any type here (external lib has no types) |
@codesyncer-inference | AI inferred something with rationale | // @codesyncer-inference: Page size 20 (standard UX) |
@codesyncer-decision | Post-discussion decision | // @codesyncer-decision: [2024-10-15] Using Stripe (intl payment) |
@codesyncer-todo | Needs user confirmation | // @codesyncer-todo: Confirm API endpoint URL |
@codesyncer-context | Business context explanation | // @codesyncer-context: GDPR compliance (30-day retention) |
Existing @claude-* tags are fully compatible:
@claude-rule = @codesyncer-rule @claude-inference = @codesyncer-inference @claude-decision = @codesyncer-decision @claude-todo = @codesyncer-todo @claude-context = @codesyncer-contextCodeSyncer automatically pauses AI work when critical keywords are detected, preventing costly mistakes.
- 💰 Payment & Billing: payment, billing, subscription, charge, refund
- 🔐 Auth & Security: authentication, login, permission, encrypt, token, jwt
- 🗑️ Data Operations: delete, remove, drop, migrate, schema change
- 📜 Privacy & Compliance: personal data, GDPR, privacy, PII
- AI detects keyword (e.g., "payment")
- Automatically pauses work
- Presents recommendation + alternatives
- Waits for your decision
- Records decision in
DECISIONS.md+ code comments - Resumes work
CodeSyncer fully supports both Korean and English:
- Installation prompts
- Generated documentation
- Comment guidelines
- All UI messages
Switch language anytime during setup or use language-specific commands.
CodeSyncer automatically detects your project type and tech stack:
Supported:
- ☕ Java (Spring Boot)
- 🐍 Python (Django, FastAPI)
- 📘 TypeScript / JavaScript
- ⚛️ React / Next.js
- 🟢 Node.js / Express
- 📱 React Native
Detection happens automatically by scanning:
package.json,pom.xml,build.gradle,requirements.txt- Folder structure and dependency lists
User: "Add Stripe payment integration" AI: ⚠️ 'payment' keyword detected. Discussion needed. 💡 Recommendation: Use Stripe 🔄 Alternatives: A. Iamport (Korea only) B. Toss Payments C. Custom implementation How would you like to proceed? User: "Proceed" AI: ✅ Recording decision... Creating: - backend/src/services/PaymentService.ts /** * @codesyncer-decision: [2024-11-12] Using Stripe (international support) * @codesyncer-context: Support USD, EUR, KRW */ - frontend/src/components/PaymentForm.tsx - DECISIONS.md updated ✅ Payment integration complete!- AI makes assumptions about critical business logic
- No record of why decisions were made
- Lost context switching between repos
- Inconsistent coding patterns across team
- AI pauses for important decisions
- All decisions permanently recorded
- Seamless multi-repo workflows
- Consistent collaboration system
- Onboarding new team members takes minutes
- Claude Code (Full support)
- Cursor
- GitHub Copilot
- Continue.dev
- Codeium
Want to add support for your favorite AI tool? Contribute here!
After running codesyncer init, your project will look like:
my-project/ ├── CLAUDE.md # Claude reads this first └── .claude/ ├── CLAUDE.md # Coding guidelines ├── COMMENT_GUIDE.md # Tag usage guide ├── ARCHITECTURE.md # Project structure └── DECISIONS.md # Decision log workspace/ ├── CLAUDE.md # Claude reads this first ├── .codesyncer/ │ └── MASTER_CODESYNCER.md # Multi-repo auto-switching guide ├── backend/ │ └── .claude/ │ ├── CLAUDE.md # Coding guidelines │ ├── COMMENT_GUIDE.md # Tag usage guide │ ├── ARCHITECTURE.md # Project structure │ └── DECISIONS.md # Decision log ├── frontend/ │ └── .claude/ │ └── (same files) └── mobile/ └── .claude/ └── (same files) monorepo/ ├── CLAUDE.md # Claude reads this first ├── .codesyncer/ │ └── MASTER_CODESYNCER.md # Package navigation guide ├── packages/ │ ├── shared/ │ │ └── .claude/ │ │ └── (same files) │ └── ui/ │ └── .claude/ │ └── (same files) └── apps/ ├── web/ │ └── .claude/ │ └── (same files) └── api/ └── .claude/ └── (same files) Supported Monorepo Tools:
- ✅ Turborepo (
turbo.json) - ✅ pnpm (
pnpm-workspace.yaml) - ✅ Nx (
nx.json) - ✅ Lerna (
lerna.json) - ✅ npm/Yarn workspaces (
package.jsonwithworkspacesfield) - ✅ Rush (
rush.json)
In Expert setup mode, you can add custom keywords:
codesyncer init --mode expertThen select "Add custom keywords" and specify:
- Keywords to detect
- Severity level (CRITICAL/IMPORTANT/MINOR)
- Description
Run codesyncer update to:
- Refresh project structure in
ARCHITECTURE.md - Update comment tag statistics
- Rescan file structure
Find all tagged comments in your codebase:
# All inferences grep -r "@codesyncer-inference" ./src # All TODOs grep -r "@codesyncer-todo" ./src # All decisions grep -r "@codesyncer-decision" ./srcWe welcome contributions! CodeSyncer is open source and community-driven.
- Fork this repository
- Clone your fork:
git clone https://github.com/YOUR_USERNAME/codesyncer.git - Create a branch:
git checkout -b feature/amazing-feature - Make changes and commit:
git commit -m "feat: Add amazing feature" - Push to your fork:
git push origin feature/amazing-feature - Open a Pull Request on GitHub
- 🤖 Add support for more AI tools (Cursor, Copilot, Continue.dev)
- 🌐 Additional language translations (Japanese, Chinese, Spanish)
- 📦 More tech stack templates (Go, Rust, Ruby, PHP)
- 📝 Documentation improvements
- 🐛 Bug fixes
See CONTRIBUTING.md for detailed contribution guidelines.
- 📝 Open an Issue
- 💡 Start a Discussion
Commons Clause License + MIT
- ✅ Free to use for personal and non-commercial projects
- ✅ Free to fork and modify the code
- ✅ Free to contribute back to the project
- ❌ Cannot sell this software or provide it as a paid service
See LICENSE file for full details.
Why Commons Clause? We want CodeSyncer to remain free and accessible to all developers while preventing commercial exploitation. If you need a commercial license, please contact us.
Q: Does this only work with Claude Code? A: Currently, yes. But we're building support for Cursor, GitHub Copilot, and other tools. Contributions welcome!
Q: Can I use this on a single repository? A: Yes! CodeSyncer automatically detects if you're in a single repo (has package.json, .git, etc.) and creates .claude/SETUP_GUIDE.md instead of the multi-repo structure.
Q: Does this work with monorepos (Turborepo, pnpm, Nx, Lerna)? A: Yes! As of v2.4.0, CodeSyncer automatically detects monorepo configurations (turbo.json, pnpm-workspace.yaml, nx.json, lerna.json, or package.json with workspaces) and scans all packages in your workspace patterns.
Q: Will this slow down AI responses? A: No. CodeSyncer only adds documentation files that AI reads once per session. It actually makes AI more efficient by providing context upfront.
Q: Can I customize the keyword detection? A: Yes, use Expert setup mode to fully customize which keywords trigger discussions.
Q: Is my code/data sent anywhere? A: No. CodeSyncer is a purely local CLI tool that generates documentation files in your repos. Nothing is sent to external servers.
If CodeSyncer helps your team, please:
- ⭐ Star this repo
- 🐦 Share on Twitter
- 📝 Write about your experience
- 🤝 Contribute improvements
If you'd like to support the development of CodeSyncer, you can donate via cryptocurrency:
Ethereum (ETH) / ERC-20 Tokens:
0x0a12177c448778a37Fa4EeA57d33D06713F200De Your support helps maintain and improve CodeSyncer! 🙏
- Issues: GitHub Issues
- Discussions: GitHub Discussions
Built with ❤️ by the CodeSyncer community
Making AI collaboration seamless, one repo at a time.