AnyDoc
Rust library published by Firecrawl under the MIT licence that turns Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV and PDF into clean Markdown. It ships as a CLI, a Node.js package, a Python package and a WebAssembly build for the browser, and as an Agent Skill so Claude Code knows how to call the CLI on its own. Every format goes through one document model and one Markdown serializer: headings, numbered lists, tables with merged cells, footnotes and speaker notes come out the same way, whatever the source file.
Strengths
- One tool for Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV and text PDFs, with the same output for all
- Ships as an Agent Skill: one command and Claude Code knows when and how to call it
- Keeps the structure a model needs: tables with merged cells, numbered lists, speaker notes
- MIT licence and a local library: the document never leaves your machine
Limitations
- Does not read scans: an image page returns an error, and the OCR on offer goes through Firecrawl's hosted service
- Recent project: the performance figures on show come from the repo itself, check them on your own files
Best for
- Handing Claude Code a Word spec or an Excel sheet without losing its structure
- Preparing a set of office files before chunking and indexing them for document search