doc2md

License: MIT Docker PyPI

An MCP server that converts PDF, DOCX, and PPTX files into clean Markdown files on disk — ready for use as resources in other MCP servers.

Why This Exists

LLM toolchains, RAG pipelines, and MCP servers work best with plain text and Markdown. But organizations have years of documents locked in binary formats — PDFs, Word docs, and PowerPoint decks. doc2md bridges the gap by normalizing everything to .md files on the filesystem, so they can be immediately referenced as MCP resources, ingested into vector databases, or added to knowledge bases without any manual cleanup.

Quick Start

pip / uv

# Install with pip
pip install doc2md

# Or with uv
uvx doc2md

Docker

docker run -v $(pwd):/data benguy1000/doc2md

MCP Client Configuration

Claude Desktop (local Python)

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "doc2md": {
      "command": "python",
      "args": ["-m", "doc2md"],
      "env": {
        "TRANSPORT": "stdio"
      }
    }
  }
}

Claude Desktop (Docker)

If you run Claude Desktop with Docker-based MCP servers, mount your files directory so the container can access them:

{
  "mcpServers": {
    "doc2md": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "-v", "/path/to/your/files:/data",
        "benguy1000/doc2md"
      ]
    }
  }
}

Note: In sandboxed or Docker environments, host file paths (e.g. /mnt/user-data/uploads/) may not be accessible inside the container. If a file path doesn't resolve, use base64_content and file_name parameters instead — the LLM will automatically switch to this approach when it receives a path-not-found diagnostic.

Claude Code

Add to your project's .mcp.json or global MCP config:

{
  "mcpServers": {
    "doc2md": {
      "command": "python",
      "args": ["-m", "doc2md"]
    }
  }
}

Working directory: Claude Code launches MCP servers with your project directory as CWD. When using base64_content without an output_dir, the converted .md file is written to your project directory. Use output_dir to control where output lands.

Cursor

{
  "mcpServers": {
    "doc2md": {
      "command": "python",
      "args": ["-m", "doc2md"],
      "transport": "stdio"
    }
  }
}

Generic MCP Client (Docker + SSE)

{
  "mcpServers": {
    "doc2md": {
      "url": "http://localhost:3000/sse",
      "transport": "sse"
    }
  }
}

Tools

convert_pdf_to_markdown

Convert a PDF file to structured Markdown. Preserves headings, paragraphs, lists, tables, and page breaks.

Input:

{
  "file_path": "/path/to/report.pdf",
  "output_dir": "/path/to/output",
  "overwrite": false
}

Output:

{
  "success": true,
  "output_path": "/path/to/output/report.md",
  "file_name": "report.md",
  "source_file": "report.pdf",
  "metadata": {
    "source_format": "pdf",
    "page_count": 12,
    "word_count": 4500,
    "has_images": true,
    "image_count": 3,
    "conversion_warnings": []
  }
}

convert_docx_to_markdown

Convert a DOCX file to Markdown. Preserves headings, formatting (bold, italic, strikethrough), tables, lists, hyperlinks, footnotes, and comments.

Input:

{
  "file_path": "/path/to/document.docx",
  "output_dir": "/path/to/output"
}

Output:

{
  "success": true,
  "output_path": "/path/to/output/document.md",
  "file_name": "document.md",
  "source_file": "document.docx",
  "metadata": {
    "source_format": "docx",
    "page_count": 1,
    "word_count": 2300,
    "has_images": false,
    "image_count": 0,
    "conversion_warnings": []
  }
}

convert_pptx_to_markdown

Convert a PPTX file to Markdown. Each slide becomes an H2 section with title, body text, tables, and speaker notes.

Input:

{
  "file_path": "/path/to/deck.pptx",
  "output_dir": "/path/to/output"
}

Output:

{
  "success": true,
  "output_path": "/path/to/output/deck.md",
  "file_name": "deck.md",
  "source_file": "deck.pptx",
  "metadata": {
    "source_format": "pptx",
    "slide_count": 20,
    "word_count": 1800,
    "has_images": true,
    "image_count": 5,
    "conversion_warnings": []
  }
}

convert_auto

Auto-detect file type from extension or MIME type and route to the correct converter.

Input:

{
  "file_path": "/path/to/any-document.pdf",
  "output_dir": "/path/to/output"
}

batch_convert

Convert multiple files at once. Continues on individual failures.

Input:

{
  "file_paths": [
    "/path/to/report.pdf",
    "/path/to/notes.docx",
    "/path/to/slides.pptx"
  ],
  "output_dir": "/path/to/output"
}

Output:

{
  "results": [
    { "success": true, "output_path": "/path/to/output/report.md", "..." : "..." },
    { "success": true, "output_path": "/path/to/output/notes.md", "..." : "..." },
    { "success": true, "output_path": "/path/to/output/slides.md", "..." : "..." }
  ],
  "total": 3,
  "successful": 3,
  "failed": 0
}

Base64 Input (all single-file tools)

All single-file tools also accept base64-encoded content instead of a file path:

{
  "base64_content": "JVBERi0xLjQK...",
  "file_name": "report.pdf",
  "output_dir": "/path/to/output"
}

Common Workflows

Convert a folder of sales decks to Markdown for a knowledge base

# Using batch_convert with all PPTX files in a directory
python -c "
from doc2md.tools.batch import batch_convert
from pathlib import Path

files = [str(f) for f in Path('/data/sales-decks').glob('*.pptx')]
result = batch_convert(files, output_dir='/data/knowledge-base')
print(f'Converted {result.successful}/{result.total} files')
"

Pipe converted docs into a vector database

1. Use doc2md to convert all documents to .md files
2. Point your RAG pipeline's document loader at the output directory
3. The clean Markdown with consistent heading hierarchy makes chunking reliable

Use with Claude Desktop to convert and then query documents

1. Configure doc2md as an MCP server in Claude Desktop
2. Ask Claude: "Convert /path/to/quarterly-report.pdf to markdown"
3. The .md file is written to disk and can be referenced in follow-up queries

Configuration

Environment Variable	Description	Default
TRANSPORT	Transport mode: stdio or sse	stdio
PORT	Port for SSE transport	3000
DEFAULT_OUTPUT_DIR	Default directory for output files	Source file dir

Docker Usage

Mount your files directory so doc2md can read source files and write output:

# stdio mode (for MCP clients)
docker run -v $(pwd):/data benguy1000/doc2md

# SSE mode (for network access)
docker run -p 3000:3000 -v $(pwd):/data -e TRANSPORT=sse benguy1000/doc2md

Output .md files will appear in the mounted directory on your host filesystem.

Contributing

1. Clone the repo
2. Install dev dependencies: pip install -e ".[dev]"
3. Run tests: pytest
4. Lint: ruff check src/ tests/
5. Format: ruff format src/ tests/

License

MIT — Copyright (c) 2025 benguy1000