NoteGen is a cross-platform AI note-taking application designed for users who capture fragmented ideas via text, voice, screenshots, or links and need them transformed into coherent, well-structured Markdown notes. It targets researchers, students, and knowledge workers who value privacy, local data control, and AI-assisted writing without cloud dependency. Built with Tauri and Next.js, it supports desktop and mobile platforms with a 20MB footprint and integrates RAG, MCP, and local LLMs for intelligent note processing.
The app stores all data locally in standard Markdown files, supports WebDAV, GitHub, Gitee, and S3 sync, and leverages SiliconFlow’s free models (Qwen, BGE-M3, GLM-4V) for AI functions. It uses Tiptap/Vditor for WYSIWYG editing and includes built-in RAG, Skills, and MCP for advanced AI agent workflows—all without ads or hidden fees.
What You Get
- Local-first Markdown storage - All notes are saved as plain Markdown files on your device, with support for math formulas, mind maps, diagrams, and flowcharts using extended syntax.
- AI Agent with RAG - Uses vector embeddings (BAAI/bge-m3) and re-ranking models to retrieve relevant notes for context-aware AI responses, building a personal knowledge base from your own content.
- MCP integration - Enables AI agents to interact with external tools via Model Context Protocol, including web search, API calls, and system operations for dynamic note augmentation.
- Smart recording tools - Capture ideas via voice-to-text, screenshot OCR (using GLM-4V), image recognition, web link scraping, and PDF text extraction—all auto-saved as raw records.
- One-click note generation - AI automatically organizes fragmented records into structured, coherent Markdown notes with headings, summaries, and logical flow based on context and templates.
- Customizable Skills & Templates - Define reusable AI workflows (Skills) and templates with time-range filters and custom prompts to automate note structuring and editing tasks.
- Cross-platform sync - Synchronize notes via GitHub, GitLab, Gitee, Gitea, WebDAV, and S3—no account required to use, but full version control and backup support.
- 12+ themes with dark mode - Built-in customizable color themes and deep dark mode optimized for long reading sessions, with Tiptap/Vditor WYSIWYG editing for real-time preview.
- Free embedded AI models - Out-of-the-box access to Qwen3-8B (chat), BGE-M3 (embedding), and GLM-4.1V-9B-Thinking (vision/OCR) via SiliconFlow, with option to swap in custom models.
Common Use Cases
- Researcher compiling literature notes - A PhD student uses voice and screenshot recording to capture paper highlights, then lets AI generate structured summaries with citations from their local Markdown knowledge base.
- Developer documenting code patterns - A software engineer saves code snippets, GitHub links, and terminal outputs as records, then uses AI to auto-generate annotated Markdown docs with diagrams and context.
- Journalist gathering interview quotes - A reporter records audio interviews, extracts text, and tags them by topic; AI organizes quotes into thematic notes with embedded images and links for article drafting.
- Student preparing for exams - A university student takes quick notes during lectures via text and screenshots, then uses AI to generate flashcards and summaries from their local Markdown notes using RAG.
Under The Hood
Architecture
- Clean separation of concerns with distinct layers for UI, state management, data persistence, and native integration, enabling modularity and testability
- Centralized Zustand stores decouple UI logic from data flow, ensuring predictable state updates across complex interfaces
- Tauri-based backend abstracts system-level operations like file access and IPC, maintaining platform independence
- Drag-and-drop functionality implemented via DnD-kit with accessible, compositional wrappers that preserve component integrity
- Database layer enforces data integrity through typed interfaces and async SQL operations with normalized asset paths
Tech Stack
- Next.js 14 with React 19 and TypeScript leverages RSC and Turbopack for high-performance rendering and build optimization
- Tauri 2.x provides native desktop capabilities through plugins for filesystem, SQL, clipboard, and OS integration
- Tiptap 3.x serves as the rich-text engine with advanced extensions for math, tables, and YAML frontmatter
- Radix UI and shadcn/ui form a consistent, accessible component layer powered by Tailwind CSS and class-variance-authority
- pnpm with lockfile v9 manages a complex dependency tree including DnD Kit, Codemirror, and AntV for interactive visualizations
- Comprehensive ESLint and TypeScript presets enforce code quality with custom hooks and typing rules
Code Quality
- Extensive unit testing with precise assertions and dependency injection validates file sanitization, path resolution, and state normalization
- Modular utilities in /lib and /core emphasize pure functions with no side effects, promoting reusability and predictability
- Strong TypeScript typing ensures type safety across configurations and error handling, though custom error classes are underutilized
- Consistent component structure and prop typing enhance maintainability across mobile and desktop interfaces
- Robust UI error recovery with fallbacks and reload mechanisms, though business logic error handling remains generic
- Well-structured test and type systems imply a disciplined development workflow despite minimal explicit configuration
What Makes It Unique
- Integrated MCP server management enables automatic detection and seamless switching between stdio and HTTP-based LLMs without external tools
- Context-aware search unifies articles, notes, and tags into a single interactive interface with dynamic highlighting
- Granular model routing assigns specific LLMs to distinct user actions—chat, editing, summarization—creating purpose-built AI workflows
- Intelligent file system abstraction handles cross-platform drag-and-drop with automatic sanitization and path normalization
- Activity tracking with session-based heatmaps transforms writing patterns into visual, actionable insights
- Unified semantic component library enforces design consistency across platforms while supporting deep customization via icons and variants