An AI-assisted reader for code, Markdown, PDF, images, HTML, and .board files.
The chat backend is powered by the Claude Code SDK stack via claude-agent-sdk.
- Markdown — GFM, KaTeX math formulas, Mermaid diagrams, heading outline with scroll sync
- PDF — dual modes: native SVG rendering and Markdown reading view; region selection and text extraction
- Code — syntax highlighting for 50+ languages
- Images — auto-scaled preview
- HTML — iframe rendering with source toggle
- Board — interactive p5.js canvas with 2D/WebGL 3D drawing, animation, and interactive controls
- Three-panel layout: repository tree, document viewer, and AI chat
- Multi-repository browsing with file tree filtering
- Tabbed file viewer for multiple files simultaneously
- In-app settings for repository paths and API configuration
- Streaming conversation with Claude, powered by
claude-agent-sdk - Session history and persistence
- Context-aware tools: read visible content, selections, current PDF page, and notes
- Image upload via clipboard paste
- Canvas drawing and screenshot tools for
.boardfiles
- Sticky notes for Markdown and PDF documents
- Markdown notes anchored to heading paths and block positions
- PDF notes anchored by page and coordinate ratios
- Color options: yellow, blue, green, pink
- AI tools can read document-wide and visible notes
- p5.js drawing instructions with live rendering
- WebGL 3D support (box, sphere, lighting, camera, etc.)
- KaTeX math formula rendering on canvas
- Animation loop with configurable FPS and persistent variables
- Interactive controls: range sliders, number inputs
- Save/load board state, screenshot snapshots
git clone <repository-url> cd increa-reader pnpm run setup
setup checks prerequisites (Node.js 22+, Python 3.10+, pnpm), installs all dependencies,
creates packages/server/.venv, and generates packages/server/.env.
Edit packages/server/.env:
INCREA_REPO="/path/to/repo1:/path/to/repo2" ANTHROPIC_API_KEY="your-api-key"
You can also configure repositories and API settings from the UI settings drawer after starting the app.
To use GLM through its Anthropic-compatible API (defaults to glm-5.3):
ANTHROPIC_BASE_URL="https://open.bigmodel.cn/api/anthropic" ANTHROPIC_AUTH_TOKEN="your-glm-api-key"
pnpm dev
- Frontend:
http://localhost:5173 - Backend:
http://localhost:3000
| Command | Description |
|---|---|
pnpm dev |
Start frontend and backend concurrently |
pnpm build |
Build all packages |
pnpm check |
Run typecheck and lint |
pnpm lint |
Lint frontend code (Biome) |
pnpm lint:fix |
Lint and auto-fix |
pnpm format |
Format frontend code (Biome) |
pnpm test |
Run all tests |
Filter to a single package with --filter:
pnpm --filter @increa-reader/ui dev pnpm --filter @increa-reader/server dev
| Package | Stack |
|---|---|
packages/ui |
React 19, TypeScript, Vite (rolldown-vite), Tailwind CSS 4, shadcn/ui, Zustand |
packages/server |
FastAPI, claude-agent-sdk, PyMuPDF |
packages/pdf-reader-mcp |
Standalone MCP server for PDF reading (can be registered in Claude Code / Claude Desktop) |
Python / virtualenv issues
python3 -m venv packages/server/.venv packages/server/.venv/bin/pip install -r packages/server/requirements.txt
Chat is not available
Check that ANTHROPIC_API_KEY is set in packages/server/.env.
No repositories are shown
Check INCREA_REPO in .env, or configure from the UI settings drawer.
Backend port changed
If you change PORT, update the proxy target in packages/ui/vite.config.ts.
MIT