Any Document, Clean Text, No Cloud Needed

Something worth knowing about

Imagine you have a pile of contracts, invoices, reports — some in Word, some PDF, some Excel — and you want an AI to read all of them and do something useful. The first problem is always the same: getting the text out cleanly, quickly, and without sending sensitive files to some random third-party service.

That's exactly what anydoc solves. It's a free, open-source tool from the Firecrawl team that converts 14 document formats — Word, PowerPoint, Excel, PDFs, even old RTF files — into clean, readable text. It's fast enough to process 500 Word documents in under two seconds, and because it can run entirely on your own machine, nothing ever touches an outside server.

What makes it quietly impressive is the detail work. It figures out what kind of file it's dealing with by reading the actual file contents, not just the filename — so a renamed PDF doesn't fool it. And it works inside a browser too, which means you could build a simple tool where a client uploads a document and the AI reads it, all without that file ever being uploaded anywhere.

For founders building anything that touches documents — client onboarding, contract review, report analysis — this replaces the paid APIs that usually charge per page or per call.

Words worth knowing

Markdown — A simple way of formatting text using symbols like # for headings. AI tools read it very well.

Open-source — The code is free and public. Anyone can use it, inspect it, or improve it.

RAG pipeline — "Retrieval-Augmented Generation." Fancy phrase for: give an AI your documents so it can answer questions about them specifically.

WebAssembly — A way for software to run inside a web browser at near-native speed, without installing anything.

---

If you're currently paying to extract text from documents — or avoiding it because it felt too complicated — this one is worth a look: https://github.com/firecrawl/anydoc