ChatGPT-quality AI running 100% in your browser. No servers. No tracking. Works offline.
Try Demo → • Documentation • Discord • Twitter
💡 Replace demo.gif above with your actual demo showing: Chat working → Disconnect WiFi → Still works perfectly
Every conversation with ChatGPT trains their models. Every prompt with Claude lives on Anthropic's servers. Every question costs you $20/month.
We think AI should be different.
|
Your conversations never leave your device. Not encrypted. Not anonymized. Just... never sent. No company can see your data. No AI can train on your prompts. True privacy by architecture. |
Download once, use forever. Write on flights. Code at the beach. Work in basements. Study in libraries. No internet? No problem. |
|
No subscriptions. No API costs. No hidden fees. The models are free. The code is free. Zero cost. Forever. |
Every line of code is public. Audit it yourself. Fork it. Modify it. No trust required. Verify everything. |
# Just visit the demo
https://trylokul.com
# First time: ~30 second setup
# After that: Works forever, even offlineNo account. No download. Just works.
Normal chat, everything works smoothly
Still works perfectly. Zero internet.
| Feature | Description |
|---|---|
| 🤖 Multiple AI Models | Choose between Quick (80MB), Smart (2.8GB), or Genius (6.4GB) |
| 💬 ChatGPT-Style Interface | Familiar, clean, fast - just start chatting |
| 📦 Everything Local | Models, conversations, settings - all on your device |
| 🔄 Works Offline | After first download, internet is optional |
| 🎨 Clean UX | No clutter, just AI assistance |
| 🛡️ Privacy First | Zero data collection, zero analytics |
| ⚡ GPU Accelerated | Uses WebGPU for instant responses |
| 💾 Auto-Save | Never lose a conversation |
| 🎭 Memory System | Remembers context across long chats |
| 📱 Progressive Web App | Add to home screen, works like a native app |
| Feature | ChatGPT Plus | Lokul |
|---|---|---|
| Privacy | ❌ Cloud (your data on their servers) | ✅ 100% Local (never leaves your device) |
| Offline | ❌ No (needs internet) | ✅ Yes (works on airplanes) |
| Cost | ❌ $20/month ($240/year) | ✅ Free ($0 forever) |
| Speed | ✅ Instant (runs on your GPU) | |
| Open Source | ❌ No (black box) | ✅ Yes (audit it yourself) |
| Data Training | ❌ Your data trains their AI | ✅ Never (stays local) |
| Downtime | ✅ Never (no server to crash) |
graph LR
A[🌐 Visit Lokul] --> B[⬇️ Download AI Model]
B --> C[💾 Cache Locally]
C --> D[🎮 Run on Your GPU]
D --> E[💬 Chat Offline]
E --> F[🔒 Everything Stays Local]
style A fill:#FF6B35,stroke:#FFB84D,stroke-width:2px,color:#fff
style B fill:#FF8C42,stroke:#FFB84D,stroke-width:2px,color:#fff
style C fill:#FFB84D,stroke:#FF6B35,stroke-width:2px,color:#fff
style D fill:#FF6B35,stroke:#FFB84D,stroke-width:2px,color:#fff
style E fill:#FF8C42,stroke:#FFB84D,stroke-width:2px,color:#fff
style F fill:#10B981,stroke:#059669,stroke-width:3px,color:#fff
- 🌐 Visit Lokul - Opens in your browser (Chrome/Edge)
- ⬇️ Download Model - One-time download, cached forever
- 🎮 GPU Acceleration - Runs on your graphics card via WebGPU
- 💾 Local Storage - Everything saved in IndexedDB
✈️ Offline Ready - Works without internet after first load
Tech Stack:
- WebGPU for GPU acceleration
- WebLLM for model inference
- React 18 + Vite for blazing-fast UI
- IndexedDB for local storage
- Service Workers for offline support
Node.js 18+ (LTS recommended)
npm or pnpm
Chrome 113+ or Edge 113+ (for WebGPU support)# Clone the repository
git clone https://github.com/pouryaak/lokul.git
# Navigate to directory
cd lokul
# Install dependencies
npm install
# Start development server
npm run dev
# Open browser to http://localhost:5173# Build optimized production bundle
npm run build
# Preview production build
npm run preview
- ChatGPT-style interface
- Multiple AI models (Quick/Smart/Genius)
- Works 100% offline
- Performance monitoring
- Memory system
- Auto-save conversations
- Search conversations
- Better markdown rendering
- Keyboard shortcuts
- Export improvements
- Mobile optimization
- Document upload (PDF support)
- Custom system prompts
- Conversation folders
- Advanced memory management
- Personal AIs (custom personas)
- Project workspaces
- Image understanding
- Voice input/output
We love contributions! Whether it's:
- 🐛 Bug reports
- ✨ Feature requests
- 📝 Documentation improvements
- 🔧 Code contributions
All are welcome!
- Fork the repo
- Create a feature branch (
git checkout -b feature/AmazingFeature) - Commit changes (
git commit -m 'Add AmazingFeature') - Push to branch (
git push origin feature/AmazingFeature) - Open a Pull Request
See CONTRIBUTING.md for detailed guidelines.
Looking for a place to start? Check out our good first issues.
Found a bug? Have an idea? We want to hear it!
Join our growing community:
Lokul stands on the shoulders of giants:
- WebLLM - For making browser-based LLMs possible
- Meta - For open-sourcing Llama models
- Microsoft - For Phi models
- Mistral AI - For Mistral models
- shadcn/ui - For beautiful UI components
- Tailwind CSS - For utility-first CSS
This project is licensed under the MIT License - see the LICENSE file for details.
MIT License - You can do whatever you want with this code!
- ✅ Zero data collection - We literally can't see your data
- ✅ No analytics - Not even anonymous usage stats
- ✅ No tracking - No cookies, no fingerprinting
- ✅ Open source - Verify everything yourself
- Open browser DevTools (F12)
- Go to Network tab
- Use Lokul
- Watch: Zero network requests
Your conversations never leave your browser. We can prove it.
Is it really free? What's the catch?
Yes, actually free. No catch.
The AI models (Phi, Llama, Mistral) are open source and free to use. The code is open source and free to use. There's no server to pay for (because there is no server).
We might add optional premium features later (like cloud sync), but the core will always be free.
How good is it compared to ChatGPT?
Honest answer: It depends.
- Quick Mode: Good for simple questions, faster than ChatGPT
- Smart Mode: Comparable to ChatGPT 3.5 for most tasks
- Genius Mode: Approaching ChatGPT 4 quality
The gap is closing fast. And unlike ChatGPT:
- ✅ Works offline
- ✅ Costs $0
- ✅ Doesn't train on your data
Will this work on my device?
If you have:
- Chrome 113+ or Edge 113+
- 4GB+ available RAM
- A modern graphics card (even integrated)
Then yes, it'll work.
Try Quick Mode first (works on almost anything). Smart Mode needs more power.
Can you see my conversations?
No. And we can prove it.
- Open browser dev tools (F12)
- Go to Network tab
- Use Lokul
- Watch: Zero network requests
Your conversations never leave your browser. We literally can't see them.
Plus, we're open source. Audit the code yourself.
Is this legal?
Yes! 100% legal.
The models we use are open source and licensed for commercial use:
- Phi (Microsoft - MIT License)
- Llama (Meta - Llama License)
- Mistral (Mistral AI - Apache 2.0)
This is how open source is supposed to work.
If you find Lokul useful, consider:
- ⭐ Starring this repo - Helps others discover it
- 🐦 Sharing on Twitter - Spread the word
- 🐛 Reporting bugs - Help us improve
- 💻 Contributing code - Make it better
- 📝 Writing about it - Blog posts, videos, tutorials
Every bit helps!
Made by Pourya Ak
If you believe AI should be private, give us a star ⭐
© 2026 Lokul. Open source under MIT License.