MCP PDF to Markdown Converter and Crawler 📄➡️📝

This project provides a robust system for converting PDF documents to Markdown format and crawling web content using a Multi-Server Communication Protocol (MCP) architecture. It comprises two main modules: convert_pdf for PDF upload and conversion, and crawl_mcp for web crawling, along with a client application that orchestrates operations using a reactive agent.

Project Structure

The core components of this project are:

• convert_pdf: A FastMCP server (running on http://127.0.0.1:8001) responsible for handling PDF file uploads and converting them to Markdown. It includes two endpoints:

  • /upload/mcp/upload_pdf_tool: Handles PDF file uploads via multipart form data.

  • /mcp: Converts uploaded PDFs to Markdown using the convert_pdf_to_markdown_tool.

• crawl_mcp: A server module for crawling web content. For details on running this module, see src/crawl_mcp/README.md.

• client: A client application that acts as an intelligent agent. It uses LangChain and LangGraph to interact with the MCP servers, upload PDFs, and trigger conversions or crawling tasks.

Related MCP server: pymupdf4llm-mcp

Getting Started 🚀

Follow these steps to set up and run the project:

1. Prerequisites

• Python 3.9+

• uv: A fast Python package installer and resolver. Install it via pip if not already present:

  pip install uv

2. Project Setup

1. Clone the repository (if applicable) or navigate to your project root.

   cd /path/to/your/MCP

2. Create and Sync Virtual Environment: uv will create a .venv directory and install all necessary dependencies based on your pyproject.toml.

   uv sync

3. Activate the Virtual Environment: This ensures all commands run within your isolated environment.

   • macOS/Linux:

     source .venv/bin/activate

   • Windows (Command Prompt):

     .venv\Scripts\activate.bat

   • Windows (PowerShell):

     .venv\Scripts\Activate.ps1

4. Create .env file: Create a file named .env in the project root (MCP/) and add your Google Gemini API key:

   GEMINI_API_KEY_2="YOUR_GEMINI_API_KEY_HERE"

   Replace "YOUR_GEMINI_API_KEY_HERE" with your actual API key.

3. Running the Modules

Each module has its own setup and running instructions. Refer to the module-specific READMEs for details:

• Convert PDF Module: See src/convert_pdf/README.md for instructions on running the convert_pdf server.

• Crawl MCP Module: See src/crawl_mcp/README.md for instructions on running the crawl_mcp server.

4. Docker

The convert_pdf module can be run using Docker Compose with a single service:

• Service: mcp-convert-server (port 8001)

• Functionality: Handles PDF uploads and conversion to Markdown.

To run:

cd src/convert_pdf
docker-compose up --build -d

For crawl_mcp Docker instructions, refer to src/crawl_mcp/README.md.

5. Testing with Client

To test the modules, use the client application located in src/client/. Ensure the relevant servers are running, then execute:

uv run python src/client/*

For example, to test the convert_pdf module, ensure a PDF file (e.g., input/sample.pdf) exists in the project’s input directory and run:

uv run python src/client/test_client.py

For testing crawl_mcp, refer to its README for specific client instructions.

6. Directory Structure

MCP/
├── src/
│   ├── convert_pdf/
│   │   ├── README.md
│   │   ├── src/
│   │   │   ├── __init__.py
│   │   │   ├── convert_mcp.py
│   │   │   ├── pdf2md.py
│   │   │   └── upload_api.py
│   │   ├── uploaded/
│   │   ├── output/
│   │   ├── processed_files.json
│   │   ├── docker-compose.yml
│   │   ├── Dockerfile
│   │   ├── pyproject.toml
│   │   └── uv.lock
│   ├── crawl_mcp/
│   │   ├── README.md
│   │   └── (other module files)
│   ├── client/
│   │   ├── test_client.py
│   │   └── (other client scripts)
├── .env
└── README.md

Notes

• Ensure the .env file is correctly configured with your API key.

• The convert_pdf module handles both upload and conversion on port 8001, consolidating functionality for efficiency.

• For detailed module configurations, refer to the respective READMEs.

• If encountering issues (e.g., ClientDisconnect or import errors), check logs with:

  docker-compose logs mcp-convert-server