# 📄 PDF to Markdown (PDF to MD) — Fast, Private, Open-Source Converter

> **The ultimate open-source PDF to Markdown converter for Humans, Developers, and AI Agents.**  
> Effortlessly convert **PDFs, Word (`.docx`), Excel & CSVs (`.xlsx`, `.csv`), and YouTube URLs** into clean, structured, token-efficient Markdown (`.md`).

[![Live Web App](https://img.shields.io/badge/Live%20Web%20App-pdftomarkdown.fyi-FFB7B2.svg)](https://pdftomarkdown.fyi)
[![NPM Version](https://img.shields.io/badge/npx-pdftomarkdown-292524.svg)](https://www.npmjs.com)
[![GitHub Stars](https://img.shields.io/badge/GitHub-Stars-yellow.svg)](https://github.com/HEMANT16/pdftomarkdown)
[![License: MIT](https://img.shields.io/badge/License-MIT-E8EFE8.svg)](https://opensource.org/licenses/MIT)
[![100% In-Browser Privacy](https://img.shields.io/badge/Privacy-100%25%20Client--Side-526A52.svg)](https://pdftomarkdown.fyi)
[![AI / MCP Ready](https://img.shields.io/badge/Model%20Context%20Protocol-MCP%20Ready-EFEDF4.svg)](https://modelcontextprotocol.io)

---

## 📑 Table of Contents

- [Why PDF to Markdown?](#-why-pdf-to-markdown)
- [Architecture & Conversion Flow](#-architecture--conversion-flow)
- [Key Features](#-key-features)
- [How to Use (3 Ways)](#-how-to-use-3-ways)
  - [1. Live Web Application](#1-live-web-application-zero-install)
  - [2. Zero-Install CLI (`npx`)](#2-zero-install-cli-npx)
  - [3. AI Agent & MCP Server](#3-ai-agent--mcp-server-claude-cursor)
- [Benchmark & Comparison Matrix](#-benchmark--comparison-matrix)
- [Supported Formats](#-supported-formats)
- [CLI Reference & Options](#-cli-reference--options)
- [Programmatic API (Node / TypeScript)](#-programmatic-api-node--typescript)
- [Technical SEO & Performance](#-technical-seo--performance)
- [Frequently Asked Questions (FAQ)](#-frequently-asked-questions-faq)
- [License](#-license)

---

## 🎯 Why PDF to Markdown?

PDFs are designed for **visual printing**, not editing or AI computing. They trap content inside fixed-coordinate binary blobs. **Markdown (`.md`)** is the universal standard for:

1. 🤖 **AI & LLM Workflows**: Maximizes prompt token density for Claude, ChatGPT, Cursor, and Gemini by stripping layout bloat.
2. 🗂️ **Knowledge Management**: Effortless importing into **Obsidian, Notion, Logseq, Roam, and Apple Notes**.
3. 📝 **Technical Documentation**: Direct ingestion into static site generators like **Docusaurus, VitePress, MkDocs, and Astro**.
4. 🌿 **Git Version Control**: Clean line-by-line diffs instead of opaque binary changes.

---

## 🏗️ Architecture & Conversion Flow

```
                                  INPUT SOURCE
         ┌──────────────────────────────┬──────────────────────────────┐
         ▼                              ▼                              ▼
  📄 PDF Document                📝 Word / Excel                🎥 YouTube Link
  (Local / Remote URL)           (.docx, .xlsx, .csv)           (Video URL / Shorts)
         │                              │                              │
         └──────────────────────────────┼──────────────────────────────┘
                                        ▼
                     ┌──────────────────────────────────────┐
                     │     Unified Geometric & NLP Engine   │
                     ├──────────────────────────────────────┤
                     │  • Visual Reading Order Sorter       │
                     │  • H1–H4 Typography Hierarchy        │
                     │  • GFM Table Reconstructor           │
                     │  • Code Block & List Parser          │
                     │  • Link & Metadata Extractor         │
                     └──────────────────────────────────────┘
                                        │
                                        ▼
                                 CLEAN MARKDOWN
         ┌──────────────────────────────┼──────────────────────────────┐
         ▼                              ▼                              ▼
  🌐 Web UI Workspace           💻 CLI Output Stream           🤖 AI Agent Context
  (Split Editor + Preview)       (stdout / .md file export)     (Claude / Cursor MCP)
```

---

## ✨ Key Features

- 🔒 **100% In-Browser Privacy**: In the web app, documents are converted locally in memory via WebAssembly / JavaScript. Zero files are uploaded to any external server.
- ⚡ **Zero-Install CLI (`npx`)**: Convert any PDF, Word doc, or CSV straight from your command line with no setup.
- 📊 **Smart Table Detection**: Automatically aligns multi-column rows into clean GitHub Flavored Markdown (GFM) pipe tables (`| Header | ... |`).
- 🗂️ **Multi-Format Extensibility**: Converts PDFs, Word (`.docx`), Excel/CSVs (`.csv`), and YouTube videos into Markdown.
- 🤖 **Native Model Context Protocol (MCP) Server**: Connect directly to **Claude Desktop**, **Cursor**, **Windsurf**, and **Antigravity** with zero coding.
- 🎨 **Minimalist Dual-Panel Workspace**: Live side-by-side raw Markdown editor and rendered HTML preview with line, word, and character counters.
- 🚀 **Blazing Fast**: Converts standard multi-page documents in milliseconds without requiring heavy GPUs or multi-gigabyte models.

---

## 🚀 How to Use (3 Ways)

### 1. Live Web Application (Zero Install)

Visit **[pdftomarkdown.fyi](https://pdftomarkdown.fyi)**:
1. Drag and drop any PDF file.
2. Click **Convert to Markdown**.
3. Preview the rendered document side-by-side and click **Copy Markdown** or **Download .md**.

---

### 2. Zero-Install CLI (`npx`)

No installation required. Run directly in your terminal:

```bash
# Convert a local PDF file
npx pdftomarkdown research-paper.pdf -o paper.md

# Convert a Word document (.docx)
npx pdftomarkdown document.docx -o doc.md

# Convert a CSV table to GFM pipe tables
npx pdftomarkdown financials.csv -o financials.md

# Extract timestamped transcript from a YouTube video
npx pdftomarkdown "https://www.youtube.com/watch?v=dQw4w9WgXcQ" > video.md

# Convert with an automatic Table of Contents
npx pdftomarkdown manual.pdf --toc -o manual.md

# Pipe directly into an AI tool / LLM
cat quarterly-report.pdf | npx pdftomarkdown | llm "Summarize the key takeaways"
```

---

### 3. AI Agent & MCP Server (Claude / Cursor)

Equip your AI assistants with PDF to Markdown capabilities with **zero coding**.

#### Claude Desktop Configuration:
Add this entry to your `claude_desktop_config.json`:
```json
{
  "mcpServers": {
    "pdftomarkdown": {
      "command": "npx",
      "args": ["-y", "pdftomarkdown", "--mcp"]
    }
  }
}
```

Now you can prompt Claude:
> *"Read `whitepaper.pdf` and convert it to clean markdown with table of contents."*

---

## 📊 Benchmark & Comparison Matrix

| Capability | **pdftomarkdown** | **PSPDFKit** | **Marker** | **Microsoft MarkItDown** |
| :--- | :---: | :---: | :---: | :---: |
| **Instant Live Web App** | ✅ **Yes (pdftomarkdown.fyi)** | ❌ No | ❌ No | ❌ No |
| **100% In-Browser Privacy** | ✅ **Yes (0 uploads)** | ❌ No | ❌ No | ❌ No |
| **Zero-Install CLI (`npx`)** | ✅ **Yes** | ✅ Yes | ❌ (Heavy Python) | ❌ (Python pip) |
| **Live Rendered Preview** | ✅ **Yes (Side-by-side)** | ❌ No | ❌ No | ❌ No |
| **Word (`.docx`) to MD** | ✅ **Yes** | ❌ No | ❌ No | ✅ Yes |
| **Excel / CSV to GFM Table** | ✅ **Yes** | ❌ No | ❌ No | ✅ Yes |
| **YouTube Video to Transcript** | ✅ **Yes** | ❌ No | ❌ No | ✅ Yes |
| **Model Context Protocol (MCP)** | ✅ **Yes (Built-in)** | ⚠️ Partial | ❌ No | ⚠️ Python only |
| **Execution Speed** | ⚡ **< 200 ms / page** | ⚡ < 200 ms / page | 🐢 5–20 s (GPU) | ⚡ < 500 ms / page |
| **Hardware Overhead** | 🪶 **Featherweight (JS)** | 🪶 Featherweight | 🐘 Giant (10GB+ PyTorch) | 🐍 Python Env |

---

## 🗂️ Supported Formats

| Format | Extension | Output Style |
| :--- | :--- | :--- |
| **PDF Documents** | `.pdf` | Headings (H1–H4), lists, GFM tables, links, code blocks |
| **Word Documents** | `.docx` | Headings, bold/italic, bullet lists, clean links |
| **Excel & CSV** | `.csv`, `.tsv`, `.xlsx` | GitHub Flavored Markdown (GFM) pipe tables |
| **YouTube URLs** | `https://youtube.com/...` | Timestamped transcript (`00:00`), description, and metadata |
| **Plain Text & Code** | `.txt`, `.json`, `.xml` | Formatted code fences & readable Markdown blocks |

---

## 🛠️ CLI Reference & Options

```bash
pdftomarkdown — The universal document to Markdown converter.

Usage:
  npx pdftomarkdown <file-or-url> [options]

Options:
  -o, --output <file>    Write output directly to a file (.md)
  --toc                  Automatically generate a Table of Contents
  --no-tables            Disable automatic GFM table detection
  --mcp                  Start Model Context Protocol (MCP) server
  -v, --version          Show current version
  -h, --help             Show help documentation
```

---

## 💻 Programmatic API (Node / TypeScript)

You can import and use the converter directly inside your Node.js or TypeScript projects:

```typescript
import { convertPdfToMarkdown, convertAnyToMarkdown } from 'pdftomarkdown';

// Convert a PDF buffer or file
const markdown = await convertPdfToMarkdown(pdfBuffer, {
  includeToc: true,
  detectTables: true,
  preservePageBreaks: true,
});

console.log(markdown);
```

---

## 🔍 Technical SEO & Performance

- **Primary Keywords**: PDF to Markdown, PDF to MD, PDF to Markdown converter, convert PDF to Markdown.
- **Secondary Keywords**: PDF to Markdown online, free PDF to Markdown converter, convert PDF file to Markdown, extract Markdown from PDF, Word to Markdown, YouTube to Markdown.
- **Technical Standards**: Semantic HTML5, Schema.org `WebSite`, `WebApplication`, and `FAQPage` JSON-LD schemas, sitemap.xml, robots.txt.
- **Zero Client Overhead**: No heavy UI frameworks on the parser engine, optimized WebAssembly worker threads.

---

## ❓ Frequently Asked Questions (FAQ)

### What is PDF to Markdown conversion?
PDF to Markdown conversion parses the geometry, fonts, and text positions inside fixed-layout PDF files and translates them into plain-text Markdown (`.md`) format, preserving headings, lists, tables, and links.

### How do I convert PDF to Markdown for free?
You can use the live web app at [pdftomarkdown.fyi](https://pdftomarkdown.fyi) or run `npx pdftomarkdown input.pdf -o output.md` in your terminal. Both are 100% free with no registration or subscriptions.

### Is my document kept private?
Yes. When using the web app, all parsing occurs **locally in your browser's memory**. No files or text streams are ever sent to our servers.

### Can I convert tables from PDF to Markdown?
Yes. Our engine detects horizontally and vertically aligned text columns and formats them into standard GitHub Flavored Markdown (GFM) pipe tables.

### Can I use the output with ChatGPT, Claude, and LLMs?
Yes. Clean Markdown is the optimal input format for AI prompts and RAG vector databases, minimizing token usage while maintaining document hierarchy.

---

## 📄 License & Disclaimer

Distributed under the **MIT License**. See `LICENSE` for more information.

> **Disclaimer:** All product names, logos, and brands (including Microsoft, PSPDFKit/Nutrient, Marker, Claude, and ChatGPT) are property of their respective owners and are used here solely for identification, compatibility, and comparative purposes.

Copyright © 2026 [PDF to Markdown](https://pdftomarkdown.fyi)