Branch freely. Never start over.
A thinking tool for complex content. Ask questions, explore tangents, and always find your way back.
AI chats are single-threaded. Ask a follow-up, and your whole context gets polluted. Eventually you give up and start a new chat.
Mull fixes this. Every message can branch. Go deep into a rabbit hole, then pop back to clean context.
- 🌳 Branching conversations — Every message can spawn a thread
- 📍 Always anchored — Source content never buried
- 🔄 Clean returns — Pop back, context stays pristine
- 📝 Paste anything — URLs, markdown, HTML, plain text
- ✨ Smart highlights — Theme-aware, muted colors
- 📱 Mobile-first — Touch-native with bottom sheets
git clone https://github.com/starksama/mull.git
cd mull
pnpm install
pnpm dev📄 Source Content
└── 💬 "What does X mean?"
└── 🤖 "X means..."
├── 💬 "Tell me more" → goes deeper
└── 💬 "Back to Y" → clean context
The tree IS the data structure. Messages with parent_id. Simple.
- Next.js 16 — App Router
- Tailwind CSS 4 — Styling
- Zustand — State
- Framer Motion — Animation
- Jina Reader — Content extraction
- Content parsing (URL, paste, markdown, PDF)
- Reader UI with highlights
- Layer navigation
- Real AI integration (GPT-4o-mini)
- Magic link auth (Supabase)
- Thread persistence (localStorage)
- Export notes as markdown
- Cloud sync when logged in
- Usage tracking / billing
- Architecture — Data model
- Decisions — Why we built it this way
- Contributing — How to help
MIT
Explore freely. Return cleanly.