🧠 MCP-Based Meeting Summary Service Example

This project is a sample implementation of an AI-powered meeting summarization service using MCP (Model Context Protocol). Built with FastMCP, LangChain, PostgreSQL, and Notion. it demonstrates how to define and invoke MCP-compatible tools to generate summaries, extract keywords, and identify action items from transcribed meeting data.

🚀 Features

• Generate meeting summaries using LLMs
• Extract keywords and action items
• Modular MCP (Model Context Protocol) architecture
• Integration with Notion for storing summaries
• PostgreSQL for storing transcripts
• LangChain for LLM interactions
• FastMCP for tool management

---

⚙️ Initial Setup

1. Clone the Repository

git clone https://github.com/your-username/meeting_summary.git
cd meeting_summary

2. Create and Activate Virtual Environment

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip3 install -r requirements.txt

🧪 .env Configuration

Create a .env file in the root directory with the following content:

DATABASE_URL=postgresql://postgres:postgres@localhost:5432/meeting_summary
OPENAI_API_KEY=your_openai_api_key
NOTION_TOKEN=your_notion_token
NOTION_PAGE_ID=your_notion_page_id  # The ID of the Notion page

🐘 Start PostgreSQL (via Docker)

If you don’t have PostgreSQL running locally, use Docker Compose:

docker-compose up -d

Notion Setup

this project uses Notion to store meeting summaries. You need to create a Notion integration and get your token and page ID.

1. Go to notion-mcp-server
2. Follow the instructions to create a Notion integration and get your token.
3. Share the Notion page with your integration to allow it to write data.

▶️ Run

To test the full MCP tool pipeline using LangChain function calling:

python3 client.py

This will send a prompt to the LLM asking it to summarize the meeting and generate action items using your local MCP tool server.

📁 Folder Structure

meeting_summary/
├── mcp_server.py          # FastMCP server with tool definitions
├── client.py              # LangChain client with tool calling
├── tools/                 # MCP-compatible tools (e.g., generate_summary)
├── requirements.txt
├── .env                   # Your local environment config
├── docker-compose.yml     # PostgreSQL setup
├── init.sql               # SQL initialization script