In recent years, with the rapid advancement of LLM capabilities, AI has become deeply integrated into office scenarios. These scenarios often involve handling a wide variety of document formats - such as Word, PowerPoint, Excel, and PDF - which are not inherently LLM-friendly. Consequently, the standard approach is to convert them into Markdown, making a high-quality document converter essential.
Recently, the Firecrawl team released anydoc, a powerful, open-source, cross-platform document converter. It enables the rapid conversion of formats like Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, and PDF into clean Markdown documents.
<img src="https://media.daily.dev/image/upload/s--7BtSqFDg--/f_auto/v1786327248/ugc/content_8eef178a-8886-4227-8fad-9c58f987b41a?_a=BAMAMicg0" alt="anydoc.png">
<h3>Anydoc Features</h3>
<ul>
<li>
Built-in Agent Skills for rapid integration into various agent platforms
</li>
<li>
Fast processing speed, cross-platform, built with Rust, and provides Node.js and Python bindings.
</li>
<li>
Supports non-scanned PDF documents and runs locally, eliminating the need for OCR services.
</li>
<li>
Identifies document formats based on binary data, ensuring accurate conversion even if file extensions are incorrect.
</li>
<li>
Supports conversion across multiple document formats; features a unified document model and Markdown serializer to ensure consistent rendering.
</li>
</ul>
Currently anydoc supports the following document types:
<ul>
<li>
Word : .doc, .docx, .docm
</li>
<li>
PowerPoint : .ppt, .pps, .pot, .pptx, .pptm, .ppsx, .ppsm
</li>
<li>
Excel :  .xls, .xlsx, .xlsm, .xlsb
</li>
<li>
OpenDocument : .odt, .ods, .odp
</li>
<li>
Rich Text Format :  .rtf
</li>
<li>
EPUB :  .epub
</li>
<li>
CSV : .csv
</li>
<li>
PDF :  .pdf
</li>
</ul>
<h3>Anydoc Performance</h3>
Anydoc was compared against six mainstream converters, with testing covering 100 real-world documents across 14 formats; scores ranged from 0 to 100, with higher scores indicating better performance.
<img src="https://media.daily.dev/image/upload/s--lm9HU518--/f_auto/v1786327276/ugc/content_48bedfe9-f98c-4e6f-9143-5f1978ce0594?_a=BAMAMicg0" alt="anydoc-benchmark.png">
<h3>Anydoc Online Demo</h3>
Open <a href="https://batchtool.com/tools/docs-to-markdown" target="_blank" rel="noopener nofollow">https://batchtool.com/tools/docs-to-markdown</a> in your browser, then drag the document you wish to process into the “Drop documents, or click to choose” area to convert it.
<img src="https://media.daily.dev/image/upload/s--8iMWOd8D--/f_auto/v1786327321/ugc/content_3c62d86e-8316-4c69-b4f1-0a355f52d787?_a=BAMAMicg0" alt="batchtool-anydoc.png">
<h3>Local Deployment</h3>
CLI
<pre><code>npx @firecrawl/anydoc report.docx # Markdown to stdout

npx @firecrawl/anydoc slides.pptx -o slides.md # or to a file

npx @firecrawl/anydoc - --format csv &lt; data.csv # read stdin
</code></pre>
Node.js
1.Install firecrawl/anydoc module
<pre><code>npm install @firecrawl/anydoc

# Or

pnpm add @firecrawl/anydoc
</code></pre>
2.Using the @firecrawl/anydoc module
<pre><code>import { toDocument, toMarkdown, toMarkdownBytes } from &#x27;@firecrawl/anydoc&#x27;;

// From a file path:

const markdown = await toMarkdown(&#x27;report.docx&#x27;);

// From bytes, with the format detected from the content:

const fromBytes = await toMarkdownBytes(bytes);

// Or stop at the document model, which also carries embedded assets.

// Unsupported for `pdf`

const document = await toDocument(bytes);
</code></pre>
In real-world document scenarios, you should use <code>try…catch</code> statements to catch errors during conversion. For example, when converting a PDF file that contains only images, an exception with the error code “unsupported” will be thrown.
<img src="https://media.daily.dev/image/upload/s--BCPSVDQ8--/f_auto/v1786327340/ugc/content_9bc46b42-d7c2-40dc-8b42-6cf9ee36d919?_a=BAMAMicg0" alt="anydoc-tomarkdown-error-code.png">
anydoc error codeIf you want to use anydoc in the browser, you need to use the <code>@firecrawl/anydoc-wasm</code> module.
Agent Skill
If you want to use anydoc with Claude Code, Codex, Cursor, or OpenCode, you first need to install the built-in anydoc Agent skill:
<pre><code>npx skills add firecrawl/anydoc
</code></pre>
<h3>Summary</h3>
With the anydoc library, we can easily implement document conversion functionality across various environments or platforms. If you only need to handle PDF documents, you can use the pdf-inspector library developed by the Firecrawl team; however, it does not support image-based PDFs, so you would still need to integrate an OCR service yourself.

bytefer

The Firecrawl team released anydoc, an open-source, cross-platform document converter built with Rust that converts Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, and PDF files into clean Markdown. It offers Node.js and Python bindings, a CLI, browser support via WebAssembly, and built-in agent skills for platforms like Claude Code, Cursor, and Codex. Benchmarked against six mainstream converters across 100 real-world documents in 14 formats, anydoc claims superior performance. It detects document formats from binary data, supports local processing without OCR, and provides a unified document model for consistent Markdown output.

A Faster Open-Source Alternative to MarkItDown, Pandoc, and Docling

I don’t think native documents, misleading file extensions, and image-only PDFs belong on one speed leaderboard.