Docling MCP Server

A Model Context Protocol (MCP) server for converting documents to markdown using the Docling library. This server enables Claude and other AI assistants to process and extract content from various document formats.

Features

• Convert documents from URLs or local files to markdown
• Extract tables from documents
• Convert documents with embedded images
• Support for OCR (Optical Character Recognition)
• Batch processing of multiple documents
• Caching of conversion results for improved performance
• Hardware acceleration support (MPS on macOS)

Installation

Prerequisites

• Python 3.10 or higher
• Docling library
• MCP library

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/mcp-docling.git
   cd mcp-docling
   

2. Create a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

3. Install the package:

   pip install -e .
   

Usage

Development Mode

Run the server in development mode:

mcp dev mcp_docling/server.py

"Production" Mode (as a module)

Run the server as a module:

python -m mcp_docling

Integration with Claude Desktop

To use this server with Claude Desktop, add the following configuration to your Claude Desktop config file (located at ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "docling": {
      "command": "/path/to/your/python/environment/bin/python",
      "args": [
        "-m",
        "mcp_docling"
      ],
      "env": {
        "PYTHONPATH": "/path/to/your/project/directory"
      }
    }
  }
}

Replace the paths with your actual Python environment and project directory paths.

Available Tools

convert_document

Converts a document from a URL or local path to markdown format.

convert_document("https://arxiv.org/pdf/2408.09869")
convert_document("/path/to/document.pdf", enable_ocr=True, ocr_language=["en"])

convert_document_with_images

Converts a document and returns both markdown text and embedded images.

convert_document_with_images("https://arxiv.org/pdf/2408.09869")

extract_tables

Extracts tables from a document and returns them as structured data.

extract_tables("https://arxiv.org/pdf/2408.09869")

convert_batch

Converts multiple documents in batch mode.

convert_batch(["https://arxiv.org/pdf/2408.09869", "/path/to/document.pdf"])

get_system_info

Returns information about the system configuration and acceleration status.

get_system_info()

Testing

You can test the server using the provided test script:

python test_docling_server.py

Make sure to set the required environment variables before running the test:

export INFERENCE_MODEL="your-model-id"
export LLAMA_STACK_PORT="8080"

Configuration

The server supports various configuration options:

• OCR support with language selection
• Hardware acceleration (MPS on macOS)
• Caching of conversion results
• Batch processing settings

Troubleshooting

If you encounter issues:

1. Check the logs for error messages
2. Verify that your Python environment has all required dependencies
3. Ensure the PYTHONPATH is correctly set in your configuration
4. For hardware acceleration issues, check that your system supports the configured accelerator

License

[Your License Here]

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.