Meeting Transcript Analyzer

Meeting Transcript Analyzer - Multi-Agent MCP App

A multi-agent system that analyzes meeting transcripts using AI-powered summarization, key point extraction, and task identification. Fully compliant with the Model Context Protocol (MCP) standard.

Features

• Modern Web UI: Clean, responsive horizontal layout interface for easy interaction (see screenshot)
• Summarize: Generate concise summaries of meeting transcripts (see screenshot)
• Key Highlights: Extract and display key points as bullet points (•) (see screenshot)
• Grab Tasks: Identify actionable tasks from meeting discussions (see screenshot)
• MCP Compliant: Full adherence to Model Context Protocol standards

MCP Compliance

This application is fully compliant with the Model Context Protocol (MCP) standard as defined in the official MCP specification:

✅ MCP Standard Implementation

• Discovery Endpoint: /.well-known/mcp.json (MCP compliant)
• Tool Schema: JSON Schema format with inputSchema property
• Response Format: MCP-compliant content array with structured content
• Request Format: Standard MCP invoke endpoint with name and arguments
• Capabilities: Proper capabilities object in discovery response

🔧 MCP Endpoints

Each agent exposes MCP-compliant endpoints:

📋 MCP Tool Definitions

Summarizer Agent Tools:

• summarize_transcript: Generates concise meeting summaries
• highlight_key_points: Extracts 3-5 key insights as bullet points

Task Extractor Agent Tools:

• extract_tasks: Identifies actionable tasks from transcripts

🔄 MCP Response Format

All tools return MCP-compliant responses:

Prerequisites

• Python 3.9+
• OpenAI API key

Setup

1. Install dependencies:
   pip install -r requirements.txt
2. Configure OpenAI API Key: Create a file named openai_key.txt in the project root and add your OpenAI API key:
   sk-your-openai-api-key-here

Running the Agents

Important: All commands must be run from the project root directory.

Note: Start the sub-agents first, then the super agent to ensure proper tool registration.

1. Start the Summarizer Agent (Port 8001)

2. Start the Task Extractor Agent (Port 8002)

3. Start the Super Agent (Port 8000)

Using the Application

1. Access the Web Interface: Open your browser and go to: http://localhost:8000
2. Analyze a Transcript:
   • Paste your meeting transcript in the left textarea
   • Enter a prompt like "Summarize this meeting" or "Extract key points" in the second textarea
   • Click "Analyze Transcript"
3. View Results:
   • Results appear in the right panel with structured formatting
   • Summaries appear as formatted paragraphs
   • Key points display as clean bullet points (•)
   • Tasks show as numbered actionable items
   • Metadata shows transcript length, tool used, and point/task counts

Application Screenshots

Welcome Page

The clean, modern interface users see when first opening the application.

Summarize Flow

The application summarizing a meeting transcript with a brief, concise style.

Key Highlights Flow

Extracting key insights and main points from a meeting transcript as bullet points.

Task Extraction Flow

Identifying and extracting actionable tasks from meeting discussions.

Architecture

• Super Agent (Port 8000): Main entry point that serves the web UI and orchestrates sub-agents
• Summarizer Agent (Port 8001): Handles transcript summarization and key point extraction
• Task Extractor Agent (Port 8002): Identifies and extracts actionable tasks from transcripts

Technical Details

• Backend Formatting: All response formatting is handled by the super agent for consistent UI presentation
• Structured Responses: Responses include type, title, content, and metadata fields
• MCP Protocol: Uses Model Context Protocol for agent communication
• Responsive Design: UI adapts to mobile devices with vertical stacking

MCP Integration

For MCP Clients

To integrate with MCP-compliant clients:

1. Discover Tools:
   curl http://localhost:8001/.well-known/mcp.json curl http://localhost:8002/.well-known/mcp.json
2. Invoke Tools:
   curl -X POST http://localhost:8001/invoke \ -H "Content-Type: application/json" \ -d '{"name": "summarize_transcript", "arguments": {"transcript": "Your transcript here..."}}'

MCP Client Libraries

This application works with any MCP-compliant client library:

• MCP Python SDK
• MCP JavaScript SDK
• MCP Go SDK

For more information about the MCP standard, visit the official MCP specification and GitHub repository.

Troubleshooting

• "ModuleNotFoundError: No module named 'agents'": Make sure you're running commands from the project root directory
• "uvicorn: command not found": Use python3 -m uvicorn instead of just uvicorn
• API Key Issues: Ensure openai_key.txt exists and contains a valid OpenAI API key
• Port Conflicts: Make sure ports 8000, 8001, and 8002 are available
• MCP Discovery Issues: Verify agents are running and accessible at their respective ports

Development

Code Formatting

To maintain consistent code style, use the provided formatting script:

This will format all Python files with Black and HTML/Markdown files with Prettier.

Manual Formatting

You can also format files individually:

File Structure