Model Context Protocol (MCP) Sample Applications

This repository contains sample applications demonstrating the use of Model Context Protocol (MCP) with various integration patterns. These examples make use of the new MCP Server (Preview) feature in Azure APIM as well as demostrates how to make use of the tools being exposed via the remove MCP Server in APIM.

Projects

ConsoleMcpServerTest

A simple console application demonstrating basic MCP client functionality to connect to an MCP server and invoke tools.

ConsoleMcpServerSemanticKernelTest

This project demonstrates the integration between Model Context Protocol (MCP) and Microsoft Semantic Kernel, allowing natural language to be used to interact with MCP tools.

Getting Started with ConsoleMcpServerTest

Prerequisites

• .NET 9.0 SDK or later
• Access to an MCP server endpoint

Configuration

The application uses appsettings.json for configuration. This file is not included in the repository for security reasons.

To set up your configuration:

1. Copy the appsettings.example.json file and rename it to appsettings.json
2. Update the settings with your actual values:

{
  "PetStoreSettings": {
    "McpEndpoint": "https://your-mcp-endpoint-here/mcp",
    "SubscriptionKey": "your-subscription-key-here"
  }
}

• PetStoreSettings:
  • McpEndpoint: The URL of your MCP server endpoint
  • SubscriptionKey: Your API subscription key for the MCP service (if required)

Running the Application

Navigate to the project directory and run:

dotnet run

The application will:

1. Connect to the MCP server
2. List available tools from the server

Prerequisites

• .NET 9.0 SDK or later
• Access to an MCP server endpoint
• Access to Azure OpenAI service

Configuration

The application uses appsettings.json for configuration. This file is not included in the repository for security reasons.

To set up your configuration:

1. Copy the appsettings.example.json file and rename it to appsettings.json
2. Update the settings with your actual values:

{
  "PetStoreSettings": {
    "McpEndpoint": "https://your-mcp-endpoint-here/mcp",
    "SubscriptionKey": "your-subscription-key-here"
  },
  "AzureOpenAI": {
    "Endpoint": "https://your-azure-openai-endpoint-here/",
    "ApiKey": "your-api-key-here",
    "DeploymentName": "your-deployment-name-here"
  }
}

• PetStoreSettings:

  • McpEndpoint: The URL of your MCP server endpoint
  • SubscriptionKey: Your API subscription key for the MCP service

• AzureOpenAI:

  • Endpoint: Your Azure OpenAI service endpoint
  • ApiKey: Your Azure OpenAI API key
  • DeploymentName: The name of your Azure OpenAI model deployment (e.g., “gpt-4o”)

Running the Application

Navigate to the project directory and run:

dotnet run

The application will:

1. Connect to the MCP server
2. Initialize Semantic Kernel with Azure OpenAI
3. Load MCP tools as Semantic Kernel plugins
4. Execute a natural language request to add a pet to the store

Security Notes

• The appsettings.json file is included in the .gitignore to prevent committing sensitive information
• Always keep your API keys and subscription IDs private
• Consider using environment variables or Azure Key Vault for production deployments