Lecturer-MCP

An AI agent using the Model Context Protocol (MCP) to control Microsoft PowerPoint presentations via tool calls. Built with FastMCP and pyautogui.

Features

• Accepts slide content or instructions via MCP tool calls
• Automatically opens a PowerPoint file
• Controls presentation:
  • Start presentation (F5)
  • Next/Previous slide (→, ←)
  • Go to specific slide number
  • End presentation (ESC)
• Responds to commands like:
  • "Start presentation"
  • "Go to slide 3"
  • "Next slide"
  • "Previous slide"
  • "End presentation"
• Safety checks: ensures PowerPoint is open before sending commands
• FastMCP SSE server for LLM/agent integration

Setup

1. (Recommended) Create a Python 3.13 virtual environment using uv:

   uv venv .venv --python=3.13
   uv pip install -r requirements.txt

2. Adjust PPT_PATH and PPT_FILE in server.py if needed:
   • PPT_PATH: Path to PowerPoint executable
   • PPT_FILE: Path to your .pptx file

Running the Server

uv run python server.py

• The server runs at http://127.0.0.1:8000/sse (SSE transport)

Testing

A test client is provided:

uv run python test_client.py

This will:

• List available tools
• Start the presentation
• Go to next/previous slides
• Jump to slide 3
• End the presentation

Customization

• Add or modify tools in server.py as needed
• Integrate with LLMs or other MCP clients for automation

Notes

• pyautogui simulates keyboard input; ensure PowerPoint is in the foreground
• Timing and transitions are handled with short delays for reliability