MCP PDF to Markdown Converter

MCP PDF to Markdown Converter 📄➡️📝

This project provides a robust system to convert PDF documents into Markdown format using a Multi-Server Communication Protocol (MCP) architecture. It comprises two FastMCP servers for uploading and converting files, and a client application that orchestrates these operations using a reactive agent.

Project Structure

The core components of this project are:

• upload_mcp: A FastMCP server (running on http://127.0.0.1:8000) responsible for handling PDF file uploads.
• convert_mcp: A FastMCP server (running on http://127.0.0.1:8001) responsible for converting uploaded PDFs to Markdown.
• client_convert: A client application that acts as an intelligent agent. It uses Langchain and LangGraph to interact with the MCP servers, upload PDFs, and trigger their conversion.

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. If you don't have uv installed, you can get it via pip:
  pip install uv

2. Project Setup

1. Clone the repository (if applicable) or navigate to your project root.
   cd /path/to/your/MCP/project
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 step is crucial to ensure 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 Servers and Client

You need to run the two MCP servers in separate terminal windows/tabs, and then run the client in a third.

Terminal 1: Run upload_mcp Server

Navigate to your project root (/path/to/your/MCP/) and run: