Real-world files to Markdown, in pure TypeScript.

Convert PDF, Word, Excel, PowerPoint, OpenDocument, Apple iWork, HWP, WPS, email, OneNote, Visio, ebooks, archives, and more into GitHub-Flavored Markdown.

Runs in Node.js, Cloudflare Workers, Edge runtimes, and browsers.

No Python. No native addons. No WASM. No third-party runtime dependencies.

npm install @mdgate/converters

import { toMarkdown } from '@mdgate/converters';

const markdown = await toMarkdown(bytes, {
  path: 'report.docx',
});

The input is bytes. The output is Markdown.

Try it in your browser. Files convert locally and never leave your machine.

A JavaScript application that needs to read real-world files quickly becomes a collection of unrelated parsers and runtimes.

PDF needs one solution. Word needs another. Then Excel, PowerPoint, email, Apple Pages, HWP, OneNote, Visio, ebooks, archives...

mdgate/converters puts them behind one interface:

Uint8Array → Markdown

It is designed to stay inside the JavaScript runtime your application already uses.

• Broad format coverage: Office, PDF, iWork, HWP, WPS, email, ebooks, notebooks, archives, and more
• Portable TypeScript: Node.js, Cloudflare Workers, Edge, and browser
• Local parsing: deterministic formats are parsed without a remote conversion service
• One Markdown dialect: official converters produce consistent GitHub-Flavored Markdown
• Content-based detection: many formats are identified from their bytes, not just their extension
• Composable: install every converter, one format, or your own combination
• Nested conversion: archives and container formats can pass inner files through the same reader
• Small supply chain: no third-party runtime dependencies, native addons, or WASM

If you only need one format, install that converter:

npm install @mdgate/pdf

import { toMarkdown } from '@mdgate/pdf';

const markdown = await toMarkdown(bytes);

Each format package exposes the same toMarkdown(bytes) model.

import { create } from '@mdgate/core';
import { pdf } from '@mdgate/pdf';
import { docx } from '@mdgate/docx';
import { xlsx } from '@mdgate/xlsx';

const toMarkdown = create([
  pdf(),
  docx(),
  xlsx(),
]);

const markdown = await toMarkdown(bytes);

@mdgate/converters is create(all()). The optional path is a format hint, never a disk read. Archives pass inner files through the same reader. To add a format, implement Converter from @mdgate/core.

* Needs a callback: vision for raster images (image()), transcription for audio (audio()), video understanding for video(). Not registered by all(). SVG converts locally.

Raster images, audio, and video use callbacks, so you can plug in the vision or transcription model you already use. See @mdgate/ai for an OpenAI-compatible adapter.

bun install          # also enables .githooks/pre-commit (lint + test)
bun test
bun run dev:demo

A committed fixture corpus under test/fixtures/ is snapshot-tested. test/robustness.test.ts mutation-tests fixtures, and test/abuse.test.ts checks that hostile files still convert.

The converters are original TypeScript. The projects below were references during development, not runtime dependencies.

• firecrawl/anydoc: early work started as a TypeScript port of this library, then diverged.
• mozilla/pdf.js: layout of PDF standard encoding tables.
• BurntSushi/rust-csv: CSV record state machine.
• mdsteele/rust-cfb: OLE compound-file FAT checks.
• ashtuchkin/iconv-lite: single-byte encoding tables.
• adobe-type-tools/agl-aglfn: Adobe Glyph List for PDF glyph names.
• adobe-type-tools/cmap-resources: CMap files for PDF CID encodings.
• adobe-type-tools/mapping-resources-pdf: CID-to-Unicode maps for PDF text.
• Unicode UCD: Equivalent Unified Ideograph mappings.

MIT