MCP Docling Server logo

MCP Docling Server

by zanetworker

The MCP Docling Server provides document processing capabilities using the Docling library. It allows you to convert documents, extract tables, and generate Q&A documents.

View on GitHub

Last updated: N/A

What is MCP Docling Server?

The MCP Docling Server is a server that exposes document processing tools built on the Docling library. It allows users to perform operations such as converting documents to markdown, extracting tables, and generating question-answer pairs from documents.

How to use MCP Docling Server?

The server can be installed using pip and started using the mcp-server-lls command. You can choose between stdio and SSE transport. The server exposes several tools that can be accessed by sending requests with the appropriate parameters, such as document source, OCR settings, and language codes. It can also be integrated with Llama Stack.

Key features of MCP Docling Server

  • Convert documents to markdown format

  • Extract tables from documents

  • Generate Q&A documents

  • Support for OCR (Optical Character Recognition)

  • Batch processing of multiple documents

  • Integration with Llama Stack

Use cases of MCP Docling Server

  • Automated document conversion to markdown

  • Extracting structured data from documents

  • Building question-answering systems from document content

  • Processing scanned documents using OCR

  • Integrating document processing into LLM applications

FAQ from MCP Docling Server

What is the default transport protocol?

The default transport protocol is stdio.

How do I enable OCR?

Set the enable_ocr parameter to true when calling the conversion tools.

What environment variables are required for the qna_from_document tool?

The qna_from_document tool requires WATSONX_PROJECT_ID, WATSONX_APIKEY, and WATSONX_URL environment variables to be set.

Where are processed documents cached?

Processed documents are cached in ~/.cache/mcp-docling/.

How can I integrate this server with Llama Stack?

You can register the MCP tools with your Llama Stack client and configure an agent to use them, as shown in the example code.