home / mcp / pdf mcp flow mcp server

PDF MCP Flow MCP Server

Provides an MCP server to extract Markdown, images, and tables from PDFs, with format conversion for RAG workflows.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "dublin1231-pdf_mcp_flow": {
      "command": "uv",
      "args": [
        "--directory",
        "D:/path/to/your/mcp-pdf-flow",
        "run",
        "simple-pdf"
      ],
      "env": {
        "PDF_SEARCH_PATHS": "D:\\example;E:\\example",
        "PYTHONIOENCODING": "utf-8"
      }
    }
  }
}

PDF MCP Flow is an MCP server that turns PDFs into structured Markdown, extracts images and tables, and enables seamless format conversions for RAG workflows. It is designed to work with MCP clients, letting you process documents locally or in batches while keeping outputs clean and AI-friendly.

How to use

You use this MCP server through an MCP client. Start the server configuration once, then instruct the client to perform tasks like extracting content from PDFs, converting Markdown to Word, or batch processing entire folders. The client will automatically invoke the MCP server behind the scenes, so you can focus on what you want to achieve rather than how to call low-level commands.

How to install

Prerequisites: you need Python installed (version 3.10 or later) and a system where you can run the MCP runner. The server uses a simple package manager to run locally.

1) Clone the project repository.

git clone https://github.com/Dublin1231/PDF_MCP_Flow.git
cd mcp-pdf-flow

2) Install dependencies using the package manager provided by the project.

uv sync

3) Run the service as an MCP server. The server is designed to start automatically when you configure the MCP client. You do not start it manually from the terminal after installation; configure your MCP client to launch it in the background.

4) If you need to configure the client, see the client configuration section for how to wire the local server into your MCP client. The client will call into the MCP server using the provided runtime command.

Claude Desktop configuration and local runtime

To add this tool to Claude Desktop, edit the MCP server configuration inside the Claude Desktop settings. This creates a local runtime that the client can auto-start.

Windows example path (modify to your actual local path):

{
  "mcpServers": {
    "simple-pdf": {
      "command": "uv",
      "args": [
        "--directory",
        "D:/path/to/your/mcp-pdf-flow",
        "run",
        "simple-pdf"
      ],
      "env": {
        "PYTHONIOENCODING": "utf-8",
        "PDF_SEARCH_PATHS": ""
      }
    }
  }
}

Notes and practical tips

Environment requirements include an OS that can support Word/WPS conversions if you plan to use the Word/PDF conversion features. The server is designed to run locally with a lightweight configuration in Claude Desktop, so you can start processing PDFs without changing your existing workflow.

If you use batch processing, you can target a directory and let the server export Markdown, images, and optional tables according to your settings. The outputs keep the source directory structure unless you choose to flatten them.

Available tools

extract_pdf_content

Core tool to extract PDF content; supports text, images, and markdown output with options for page range and keywords.

batch_extract_pdf_content

Batch process a directory of PDFs, exporting content to Markdown and images with configurable output directories and structure preservation.

get_pdf_metadata

Fast retrieval of PDF metadata and table of contents.

convert_markdown_to_docx

Convert Markdown content to a Word (.docx) document.

convert_docx_to_pdf

Convert a Word document to PDF.

search_pdf_files

Fuzzy search for PDFs by filename within a directory.

batch_extract_tables

Batch extract tables from PDFs with advanced table cleaning and layout recovery.