markitdown-mcp-advanced 0.0.1

Advanced MarkItDown MCP server with PaddleOCR support for PDF, images, and Office documents

MarkItDown MCP Advanced

A powerful MCP (Model Context Protocol) server that converts various document formats to Markdown with PaddleOCR support.

Features

• Multi-format Support: PDF, images, Office documents, HTML, CSV
• OCR Integration: High-accuracy text recognition via PaddleOCR API
• URL Support: Direct processing of remote file URLs
• MCP Protocol: Full compliance with MCP standard (STDIO and HTTP modes)
• Lightweight: Core features use only Python standard library

Supported Formats

Category	Extensions
PDF	.pdf
Images	.png, .jpg, .jpeg, .gif, .bmp, .tiff, .webp
Word	.docx
PowerPoint	.pptx
Excel	.xlsx, .xls (requires extra dependency)
Web	.html, .htm
CSV	.csv

Configuration

Required environment variables:

export PADDLE_API_URL="your_api_url"
export PADDLE_TOKEN="your_token"
export MARKITDOWN_TEMP_DIR="/path/to/temp"

Usage with Claude Desktop

Add to Claude Desktop config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "markitdown": {
      "command": "markitdown-mcp",
      "env": {
        "PADDLE_API_URL": "your_api_url",
        "PADDLE_TOKEN": "your_token"
      }
    }
  }
}

Links

• GitHub Repository
• Issue Tracker

License

MIT License