MCP-Demo

yosmanovich/MCP-Demo

3.1

If you are the rightful owner of MCP-Demo and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

The Model Context Protocol (MCP) Server is a .NET implementation that provides an API for AI models to request additional context during inference, supporting both local and web-based deployments.

Tools
3
Resources
0
Prompts
0

Model Context Protocol (MCP) Server and AI Client - .NET Implementation

This project contains two .NET app implementations of a Model Context Protocol (MCP) server; one runs in standard IO mode and another runs through a web application. The web application may be deployed to Azure App Service.

The MCP server provides an API that follows the Model Context Protocol specification, allowing AI models to request additional context during inference.

The MCP tools are in a separate project. The tools are not complex; they show the basic struture of how a tool is developed and to provide a starting point for building additional tools. There is also a Tests project that demonstrates how to unit test each tool.

A simple AI Client exists that may be used to make Chat calls to your Azure AI service and to an MCP Server. This client shows the basic flow between the system, user, assistant, and tools.

Key Features

  • Azure App Service integration
  • Custom tools support

Project Structure

  • .vscode/ - VS Code workspace files
    • mcp.json - The
  • src/ - Contains the main C# project files
    • AI.Client/ - The AI Client files
      • Program.cs - The entry point for the AI Client
      • AIClient.cs - The logic for the AI Client
    • MCP.Local/ - The MCP Local Server that runs via stdio
      • Program.cs - The entry point for the MCP Local Server
    • MCP.Server/ - The MCP Server that runs via HTTP/HTTPS
      • Program.cs - The entry point for the MCP Server
    • MCP.Tools/ - Contains custom tools that can be used by models via the MCP protocol
      • EchoTools.cs - Tool for simple string echo tests
      • MultiplicationTool.cs - Example tool that performs multiplication operations
      • TemperatureConverterTool.cs - Tool for converting between Celsius and Fahrenheit
    • MCP.Tools.Tests/ - The unit tests for the MCP Tools
      • EchoToolsTests.cs - Unit Tests for the Echo Tools
      • MultiplicationToolTests.cs - Unit Tests for the Multiplication Tool
      • TemperatureConverterToolTests.cs - Unit Tests for the Temperature Converter Tool

Prerequisites

Local Development

Run the Server Locally

  1. Clone this repository
  2. Navigate to the project directory 
    cd src
  3. Install required packages 
    dotnet restore
  4. Run the project: 
    dotnet run
  5. The MCP server will be available at http://localhost:5000
  6. When you're done, press Ctrl+C in the terminal to stop the app

Creating a docker image

  1. Open terminal
  2. Navigate to MCP.Local;
   cd .\src\MCP.Local\
  1. Build the docker container:
   dotnet publish /t:PublishContainer

Testing the Available Tools

The server provides these tools:

  • Multiplication: Multiply - Multiplies two numbers
  • Temperature Conversion:
    • CelsiusToFahrenheit - Converts temperature from Celsius to Fahrenheit
    • FahrenheitToCelsius - Converts temperature from Fahrenheit to Celsius

Connect to the Local MCP Server

Using VS Code - Copilot Agent Mode
  1. Add MCP Server from command palette and add the URL to your running server's HTTP endpoint: 
    http://localhost:5000
  2. List MCP Servers from command palette and start the server
  3. In Copilot chat agent mode, enter a prompt to trigger the tool: 
    Multiply 3423 and 5465
  4. When prompted to run the tool, consent by clicking Continue

You can ask things like:

  • Convert 25 degrees Celsius to Fahrenheit
Using MCP Inspector
  1. In a new terminal window, install and run MCP Inspector: 
    npx @modelcontextprotocol/inspector
  2. CTRL+click the URL displayed by the app (e.g. http://localhost:5173/#resources)
  3. Set the transport type to HTTP
  4. Set the URL to your running server's HTTP endpoint and Connect: 
    http://localhost:5000
  5. List Tools, click on a tool, and Run Tool

Deploy to Azure

  1. Login to Azure:

    azd auth login

  2. Initialize your environment:

    azd env new

  3. Deploy the application:

    azd up

    This will:

    • Build the .NET application
    • Provision Azure resources defined in the Bicep templates
    • Deploy the application to Azure App Service

Connect to Remote MCP Server

Using MCP Inspector

Use the web app's URL:

https://<webappname>.azurewebsites.net
Using VS Code - GitHub Copilot

Follow the same process as with the local app, but use your App Service URL:

https://<webappname>.azurewebsites.net

Run the AI Client

Modify the appsettings.json
  1. Set the AZURE_OPENAI ENDPOINT to your Azure Open AI endpoint
  2. Set the AZURE_OPENAI MODEL to a model you have deployed in Azure Open AI
  3. Set the MCP_SERVER URL to the MCP server that you are running
Modify the local secrets
  1. Set the AZURE_OPENAI API_KEY to your Azure Open AI API Key
Run the AI Client

Start the AI.Client project

dotnet run --project .\src\AI.Client\AI.Client.csproj

The client will connect to your MCP server and add all tools, notifying you in the console.

You can ask the LLM to provide information; if you provide a message that can be handled by one of the tools it will call the tool to get information from the tool and report that to you.

The application will run until you either type exit and hit enter or if you press enter with no other input.

You may clear the chat history by typing

clear

You can review the sytem prompt by typing

show prompt

You can change the sytem prompt by typing

set prompt
``` or

change prompt

If you enter an empty prompt, the tool will reset the prompt back to the default prompt / what was entered in appsettings.json.

Observe that the if the LLM has knowledge that conflicts with the tools, it will notify the user of that. This may be corrected via changes to the prompt, for example: 
- "You are a math expert..."
- "You do not understand math..."
- "You are a facilitator..."

Additionally, if you modify the system prompt with calls to the tools; the AI Client should handle those tool calls as well as part of the conversation, for example: 
- "Always convert the answer from the tool's response from fahrenheit to celsius at the end."

## Clean up resources

When you're done working with your app and related resources, you can use this command to delete the function app and its related resources from Azure and avoid incurring any further costs:

```shell
azd down

Custom Tools

The project includes several sample tools in the MCP.Tools project:

  • EchoTool.cs - Performs basic string echo operations
  • MultiplicationTool.cs - Performs multiplication operations
  • TemperatureConverterTool.cs - Converts between Celsius and Fahrenheit

To add new tools:

  1. Create a new class in the Tools directory
  2. Implement the MCP tool interface
  3. Register the tool in Program.cs

References

This project is an amalgamation of several projects found on the web. Each did one thing right and this is meant to be an all-in-one, easy to consume project that discusses getting started with MCP all the way through deploying something to Azure and how it ties into both Copilot and a custom AI Client.