Local-first Markdown reader optimized for massive documents and AI
readMD, from Natsummerance, is a local-first Markdown reader and editor built to manage very large local document collections. It renders and edits Markdown while exposing a Model Context Protocol server so external AI assistants can access and analyze local documentation. The app includes semantic pagination for massive files, non-destructive rendering correction, one-click Office/PDF-to-Markdown conversion, and offline OCR. It targets researchers, developers, and students who maintain large knowledge bases and need fast, private document access with AI-assisted search.
What tasks can you actually use it for?
The app is built for reading and editing very large Markdown collections and for academic authoring. It supplies LaTeX-focused tools such as BibTeX citation cards and theorem/proof boxes, plus Office, PDF, and web-to-Markdown conversion. Offline OCR extracts text from scanned documents. A desktop companion system with Live2D and 73 micro-breathing sprites provides an always-on assistant window for note review and lightweight interactions.
How reliable are rendered views and large-file performance?
The app uses semantic pagination to handle massive files, specified for documents exceeding 8,000 lines or 500KB, so scrolling and navigation remain responsive. It applies a non-destructive rendering correction layer that fixes broken tables and unclosed formulas in memory; the original source file remains untouched unless the user explicitly saves edits. That combination addresses the kinds of freezes that massive Markdown files typically cause in other editors.
What file formats and platforms does it accept locally?
Conversion tools import Office formats DOCX, PPTX, XLSX, PDFs, and web pages into Markdown with a single action, and the app includes high-fidelity PDF text recovery. OCR runs offline through native Windows WinRT or Apple Vision engines. Platform support covers Windows 10/11, macOS 13+ on Intel and Apple Silicon, and Linux via AppImage and Deb packages, plus Chinese domestic systems like UOS and Kylin. A portable Windows build is available.
Does it integrate with local AI assistants while keeping data on-device?
The app runs a FastMCP stdio server that exposes 21 specialized tools, turning local notes into a live context source for compatible agents to read, search, and interact with documentation on the machine. The local-first architecture ensures processing stays on-device and avoids automatic cloud uploads, making the tool suitable for sensitive research notes and private knowledge bases while enabling AI-assisted workflows.
Practical judgment and recommended workflow
readMD is a practical choice for technical writers, researchers, and developers who need dependable handling of extensive notes and context for assisted drafting. Treat generated summaries as research drafts: verify factual claims against original files and use the tool to surface passages rather than accept summaries verbatim. For large archives, split work into focused repositories to keep navigation and search predictable.
