anydoc's core API is just three things, identical across all five bindings:

API	What it does	Returns
toMarkdown(path)	Path-based conversion	String
toMarkdownBytes(bytes)	Bytes-based conversion	String
toDocument(bytes)	Stop at the document model	Structured Document (with embedded asset bytes)

"From a path or from bytes" — and if you need embedded assets (images, etc.), stop at toDocument.

The five bindings

Rust

use anydoc::ConvertError;

// From a path
let markdown = anydoc::to_markdown("report.docx")?;

// From bytes
let bytes = std::fs::read("slides.pptx")?;
let markdown = anydoc::to_markdown_bytes(&bytes, None)?;

// Document model (keeps embedded assets)
let document = anydoc::to_document(&bytes, None)?;

Install: cargo add anydoc

Node.js

import { toMarkdown, toMarkdownBytes, toDocument } from '@firecrawl/anydoc'

const markdown = await toMarkdown('report.docx')
const fromBytes = await toMarkdownBytes(bytes)
const document = await toDocument(bytes)
// Error codes: error.code ('unsupported' | 'malformed' | 'encrypted' | ...)

Install: npm install @firecrawl/anydoc. Conversion runs on the libuv thread pool without blocking the event loop.

Python

import anydoc

markdown = anydoc.to_markdown("report.docx")
markdown = anydoc.to_markdown_bytes(data)
document = anydoc.to_document(data)
# Errors: exception subclasses, e.g. anydoc.UnsupportedError

Install: pip install firecrawl-anydoc. The GIL is released during conversion, so it plays well with asyncio and threads.

WebAssembly (browser)

import init, { toMarkdownBytes, toDocument } from '@firecrawl/anydoc-wasm'

await init()          // load the wasm once
const markdown = toMarkdownBytes(bytes)
const document = toDocument(bytes)

Install: npm install @firecrawl/anydoc-wasm. Conversion happens entirely in the browser — files never leave the device. The converter on this site is exactly this package.

CLI

npx @firecrawl/anydoc report.docx                # print to stdout
npx @firecrawl/anydoc slides.pptx -o slides.md   # write to a file
npx @firecrawl/anydoc - --format csv < data.csv  # read from stdin

Agent Skill

npx skills add firecrawl/anydoc

Once installed, skills-protocol agents (Claude Code, Codex, Cursor, etc.) can read office documents in your project directly.

Error handling

All languages share the same ConvertError semantics:

Variant	Meaning	What to check
unsupported	Unknown format / scanned PDF	Verify the real format; scanned PDFs need OCR (e.g. Firecrawl Parse)
malformed	Corrupt or structurally unusable	Re-save with the original application
encrypted	Password-protected	Remove the password first
resourceLimit	Crossed a safety limit (size/depth/nodes)	Split the file; confirm it isn't a zip bomb
missingPart	Required part absent from the package	The file is incomplete — re-export
io	Underlying I/O error (native only)	Check path, permissions, file locks

Node.js exposes it via error.code, Python via exception subclasses, Rust by matching the ConvertError enum. (The WASM build has no io — there's no filesystem.)

Working with embedded assets

let document = anydoc::to_document(&bytes, None)?;
// Raw bytes and media types of images/objects are on the document model
// In Markdown, those positions render as alt text

Need the images on disk? Walk the model's asset nodes and write each one by its media type. This is the standard companion step in any docx-to-md pipeline.

Things to keep in mind

• Headers, footers, page numbers, and date/time placeholders are excluded in all formats — by design, not a bug
• Speaker notes are always kept — PPT speaker notes appear in the Markdown output
• External URL images become regular Markdown images; embedded images render as alt text plus model assets
• Version: the official README wins; this page tracks v0.1.8

Go deeper per language: Node.js · Python · Rust · WebAssembly · Agent Skill. Hit an error? See Error Handling & Limits.

Contributors: souvc