Daily Delta · the full list

anydoc: get any office file into plain text an AI can read

Thanks for commenting "DOCS" — here's everything from the post in one place: what it does, how to run it four different ways, and where it falls short.

What it is

anydoc is a free, open-source tool from Firecrawl that turns Word, PowerPoint, Excel, and 11 other document formats into clean Markdown — plain text that an AI model can actually read. Most AI tools only understand text; a raw .docx or .pdf is a binary file with none of the words exposed, so something has to pull the text out first. anydoc does that in a few milliseconds, without sending your file to any server.

Supported formats

Type	Extensions
Word	.doc .docx .docm
PowerPoint	.ppt .pps .pot .pptx .pptm .ppsx .ppsm
Excel	.xls .xlsx .xlsm .xlsb
OpenDocument	.odt .ods .odp
Other	.rtf .epub .csv .pdf

Try it with no install

Drop a file into the browser demo — it converts locally, nothing leaves your machine: firecrawl.github.io/anydoc

Run it from the command line

No install needed — this downloads and runs it once:

Terminal

npx @firecrawl/anydoc report.docx

Use it in Node.js

Install

npm install @firecrawl/anydoc

Code

import { toMarkdown } from '@firecrawl/anydoc';
const markdown = await toMarkdown('report.docx');

Use it in Python

Install

pip install firecrawl-anydoc

Code

import anydoc
markdown = anydoc.to_markdown("report.docx")

Give it to your AI coding assistant

anydoc ships as an Agent Skill — a plug-in an AI coding assistant can call on its own. One command and Claude Code, Cursor, or Codex can read office files without you converting them by hand first:

Terminal

npx skills add firecrawl/anydoc

Recap

Reads 14 file types — Word, Excel, PDF, and more — through one command
Free, open-source (MIT license), and runs locally — nothing uploaded
Converts most documents in a few milliseconds
Works as a CLI, a Node/Python/Rust library, in the browser, or as an AI agent skill

The catch

It can't read scanned images of documents — there's no OCR (the tech that turns a photo of text into real text) built in — and it skips password-locked files. Firecrawl's hosted Parse API covers both if you need them, for a fee.

Links

GitHub repo npm package PyPI package Rust crate