MDFlux turns documents into Markdown to cut AI token costs
MDFlux is a Windows desktop app that converts files such as PDFs, Word documents, PowerPoint decks, , ebooks, web pages, images, and audio into Markdown that is easier for AI systems to read. It can recover text from and image-based files with OCR, and it can process a whole folder at once. After the first setup, conversions run on the user’s own computer, so documents do not need to be uploaded to a cloud service.
The project says this approach uses about 2 to 6 times fewer tokens than sending document pages as images to a . In one scanned-PDF example, the route used 10,731 tokens, while MDFlux used OCR to produce Markdown with 1,893 tokens. The core conversion engine comes from Microsoft’s MarkItDown, while MDFlux adds scan handling, a desktop interface, , progress updates, cancellation, preview, and .
It currently targets Windows 10 and 11, and the first launch needs internet access to set up a local Python environment. Planned additions include an MCP server, a , macOS builds, code signing, and more OCR language support.
Key points
- MDFlux converts documents into Markdown so AI systems can read them with fewer tokens.
- It supports PDFs, Office files, ebooks, web pages, images, and audio.
- OCR lets it extract text from and image-based documents.
- The project claims about 2 to 6 times fewer tokens than sending pages to a .
- After first setup, it , which helps avoid uploading private documents.