News Compiler AI Agent with Azure AI Agent Service and MCP

This project showcases an AI-powered news compiler agent built with Azure AI Agent Service that seamlessly integrates with a Model Context Protocol (MCP) Server. The agent allows users to search for news on various topics using Azure’s Bing Search tool, summarize the findings, and catalog this information into an Azure Blob Storage account via the MCP Server.

Overview

The News Compiler AI Agent demonstrates how an Azure AI Agent Service agent can discover and utilize tools and resources exposed by an MCP Server. This architecture creates a loosely coupled system where the agent and storage operations are separate concerns, allowing for greater flexibility and maintainability.

Features

• Search for news on topics the user is interested in using the Bing Search tool
• Summarize news articles into digestible formats
• Store and organize these summaries in Azure Blob Storage by topic
• Manage storage containers and blobs through natural language commands

Architecture

The application follows a modular architecture that leverages Azure services along with the Model Context Protocol (MCP) to create a loosely coupled system.

Components

1. Azure AI Agent Service Agent:

   • Provides natural language understanding using a GPT model
   • Implements a Bing Search Tool for retrieving current news
   • Dynamically discovers and integrates with MCP Server tool actions

2. MCP Server:

   • Implements the Model Context Protocol
   • Exposes Azure Blob Storage operations as tool actions
   • Provides a schema for the agent to understand available operations

3. Microsoft Bot Framework Application:

   • Hosts the AI Agent within a Bot Framework app
   • Manages conversation state and user profiles
   • Relays messages between users and the Azure AI Agent Service

4. Azure Blob Storage:

   • Stores news summaries by topic in organized containers
   • Provides persistent storage for later retrieval

How It Works

Step 1: Creating an Agent

When initializing the system, the application connects to the MCP Server to discover available tools and resources:

1. The application uses the MCP Client to establish a connection with the MCP Server
2. It retrieves the schema of all available tools and resources from the MCP Server
3. For each tool discovered, it creates a function stub and registers it with the Azure AI Agent Service
4. The agent is configured with appropriate instructions to use these tools effectively

When the MCP Server adds or updates tools, running this step again will detect the changes and update the functions registered with the Azure AI Agent Service.

Step 2: Running the Azure AI Agent Service Assistant

Once the agent is created and configured:

1. Users interact with the agent through the Bot Framework Emulator using natural language