Stop Feeding Claude Raw Files
Drag a raw PDF into Claude and a lot of what it reads isn't your content at all — it's layout, metadata and embedded images, all costing tokens before your first question. MarkItDown, Microsoft's open-source converter, turns files into clean Markdown first, so Claude reads them for less and answers more precisely.
Convert Anything to Clean Markdown
MarkItDown strips out the formatting noise and outputs Markdown, which language models read natively. One tool handles PDF, Word, Excel and PowerPoint files, images (with OCR), audio, HTML, and even YouTube URLs.
Two Ways to Use It
- From the command line: install with
pip install 'markitdown[all]', then convert a file in one command, such asmarkitdown report.pdf > report.md. - Inside Claude: it ships an official MCP server. Connect it to Claude Desktop once, and Claude can convert files on demand instead of you pasting raw PDFs.
Set Expectations
It works best on text-based files; scanned PDFs and complex layouts give mixed results. The often-quoted "around 70% fewer tokens" is a ballpark that depends on the file.
What's in the Guide
Why raw files waste tokens, every format MarkItDown handles, the command-line and MCP setup, where it works best, and a direct link to the repo.
The full field guide is in the PDF.
Get the Guide
Drop your email below and we'll send it straight to your inbox.