Document → Markdown
Turn documents into Markdown. Without them leaving your laptop.
Drop a Word, PDF, Excel, or PowerPoint file below. It's read and parsed right here, in this browser tab — then you paste the clean Markdown into whichever model you're using, at a fraction of the tokens the original file would cost.
The process
Three steps, all in this tab
Drop a file
Choose any supported document. It's read into memory with the browser's File API — nothing is queued for upload.
Parsed in your tab
A handful of open-source libraries — Mammoth, PDF.js, SheetJS — turn the file's structure into clean Markdown, on your machine.
Copy or download
Paste the result into whichever model you're using, or save it as a plain .md file for later.
Coverage
What it reads
Five format families, each converted with a purpose-built parser instead of one catch-all guess.
Images, audio, and video aren't supported here, on purpose. Reading them well enough to matter means running a model, and a model means a server call. We'd rather leave a gap in the format list than quietly send your file somewhere to fill it.
Where the file goes
Nowhere. That's the whole point.
We load a few open-source parsing libraries once, from a public CDN. After that, your document is read straight into memory and never leaves the tab.
Most document converters
- Your file is uploaded to a server before anything happens
- Conversion depends on that server being up, and fast
- The file — and whatever's inside it — sits in someone's logs
Lokal.md
- Parsing runs in the browser tab you already have open
- Works on a train, a flight, or a locked-down network
- Nothing about the file's contents is transmitted, ever
Why bother converting first
Markup costs tokens. Words don't.
A one-page Word document is mostly XML underneath — styles, theme data, revision history — wrapped around a few hundred words. Paste that raw into a model and you're paying to have it parse markup it doesn't need. Markdown keeps the words, the headings, and the tables, and throws the rest away.
Good to know
A few honest answers
Nothing to clean up — it was never sent anywhere. Closing the tab clears it from memory, same as any other unsaved browser data.
Yes, once the page and its libraries have loaded once in a given session. There's no per-conversion network call.
Scanned PDFs — photos of pages, with no embedded text layer — have nothing for this tool to read. That's a job for OCR, and OCR is a job for a model, which is the one dependency we're avoiding here.