What is Meai Azure Ai Foundry Mcp Sample

meai-azure-ai-foundry-mcp-sample is a repository that showcases sample applications built using .NET, integrating AI capabilities through Microsoft.Extensions.AI, Azure AI Foundry, Foundry Local, and MCP server/client.

Use cases

Use cases include developing AI-infused applications, prototyping AI solutions, and demonstrating the capabilities of Azure AI Foundry and MCP server in real-world scenarios.

How to use

To use meai-azure-ai-foundry-mcp-sample, clone the repository, set up the prerequisites including .NET SDK and Azure Subscription, build the Foundry Local SDK, and configure the necessary files to run the sample applications on your local machine.

Key features

Key features include easy integration of AI into applications, support for both frontend and backend development using .NET, and the ability to run sample applications locally with Foundry Local and MCP server.

Where to use

meai-azure-ai-foundry-mcp-sample can be used in various fields such as software development, AI application development, and educational purposes for learning about AI integration in .NET applications.

MEAI, Azure AI Foundry, Foundry Local and MCP Sample

Did you know .NET is the most versatile language to build both frontend and backend apps? In addition to that, .NET is also capable of building AI infused app really easily. This repository provides several sample apps using Microsoft.Extensions.AI, Azure AI Foundry, Foundry Local and MCP server/client.

Prerequisites

• .NET 9 SDK
• Visual Studio 17.14 or later
• VS Code + C# Dev Kit
• Azure Subscription
• Foundry Local

Getting started

Build Foundry Local SDK

NOTE: This is an interim solution until the official NuGet package is ready.

1. Clone Foundry Local repository.

   git clone https://github.com/microsoft/Foundry-Local.git foundry-local
   

2. Build the C# SDK.

   cd foundry-local/sdk/cs
   dotnet restore && dotnet build
   

3. Get the NuGet package location.

   # bash/zsh
   FOUNDRY_LOCAL_NUGET_PACKAGE_PATH=$(find $(pwd) -name "*.nupkg" -type f -exec dirname {} \; | sort -u)
   

   # PowerShell
   $FOUNDRY_LOCAL_NUGET_PACKAGE_PATH = $(Get-ChildItem -Path . -Filter *.nupkg -Recurse).Directory.FullName
   

Get ready to run apps on your local machine

1. Clone this repository.

   git clone https://github.com/devkimchi/meai-azure-ai-foundry-mcp-sample.git
   

2. Create nuget.config. {{FOUNDRY_LOCAL_NUGET_PACKAGE_PATH}} is the value from the previous section.

   # bash/zsh
   cd meai-azure-ai-foundry-mcp-sample
   cat ./nuget.sample.config \
       | sed 's|/path/to/foundry-local/sdk/cs/src/bin/Debug|{{FOUNDRY_LOCAL_NUGET_PACKAGE_PATH}}|g' \
       > nuget.config
   

   # PowerShell
   cd meai-azure-ai-foundry-mcp-sample
   $(Get-Content -Path ./nuget.sample.config) `
       -replace "/path/to/foundry-local/sdk/cs/src/bin/Debug", "{{FOUNDRY_LOCAL_NUGET_PACKAGE_PATH}}" `
       | Out-File -Path nuget.config -Force
   

   NOTE: This step is only required until the official NuGet package is published.

Run Microsoft.Extensions.AI sample app

1. Make sure you have API keys for OpenAI, Azure OpenAI, GitHub Models, Anthropic and/or Google Vertex AI.

2. Store those API keys to user secrets:

   # OpenAI
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:openai "Endpoint=https://api.openai.com/v1;Key={{OPENAI_API_KEY}}"
   
   # Azure OpenAI
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:openai "Endpoint={{AZURE_OPENAI_ENDPOINT}};Key={{AZURE_OPENAI_API_KEY}}"
   
   # GitHub Models
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:openai "Endpoint=https://models.inference.ai.azure.com;Key={{GITHUB_PAT}}"
   
   # Anthropic
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:anthropic "Endpoint=https://api.anthropic.com;Key={{ANTHROPIC_API_KEY}}"
   
   # Google Vertex AI
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:google "Endpoint=https://generativelanguage.googleapis.com;Key={{GOOGLE_API_KEY}}"
   

3. Make sure which LLM you’re going to use at src/Meai.ClientApp/appsettings.json

   // To use OpenAI, Azure OpenAI or GitHub Models
   {
     "MEAI": {
       "ChatClient": "openai"
     }
   }
   

   // To use Anthropic
   {
     "MEAI": {
       "ChatClient": "anthropic"
     }
   }
   

   // To use Google Vertex AI
   {
     "MEAI": {
       "ChatClient": "google"
     }
   }
   

4. Run the app.

   dotnet watch run --project ./src/Meai.ClientApp
   

Run Azure AI Foundry sample app

1. Make sure you have an API key for Azure AI Foundry.

2. Store the API key to user secrets:

   # Azure AI Foundry
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:foundry "Endpoint={{AZURE_AI_FOUNDRY_ENDPOINT}};Key={{AZURE_AI_FOUNDRY_API_KEY}}"
   

3. Make sure which LLM you’re going to use at src/Foundry.ClientApp/appsettings.json

   {
     "MEAI": {
       "ChatClient": "foundry"
     }
   }
   

4. Run the app.

   dotnet watch run --project ./src/Foundry.ClientApp
   

Run Foundry Local sample app

1. Make sure you have installed Foundry Local CLI.

2. Add model to Foundry Local. You can add any model from the list by running foundry model ls.

   foundry model download qwen2.5-0.5b
   

3. Make sure which LLM you’re going to use at src/Foundry.ClientApp/appsettings.json

   {
     "MEAI": {
       "ChatClient": "local"
     }
   }
   

4. Run the app.

   dotnet watch run --project ./src/Foundry.ClientApp
   

Run MCP server/client sample app

1. Make sure you have API keys for OpenAI, Azure OpenAI or GitHub Models.

2. Store those API keys to user secrets:

   # OpenAI
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:openai "Endpoint=https://api.openai.com/v1;Key={{OPENAI_API_KEY}}"
   
   # Azure OpenAI
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:openai "Endpoint={{AZURE_OPENAI_ENDPOINT}};Key={{AZURE_OPENAI_API_KEY}}"
   
   # GitHub Models
   dotnet user-secrets --project ./src/Meai.ClientApp set ConnectionStrings:openai "Endpoint=https://models.inference.ai.azure.com;Key={{GITHUB_PAT}}"
   

3. Run the MCP server app.

   dotnet run --project ./src/McpTodo.ServerApp
   

4. Run the MCP client app in another terminal.

   dotnet watch run --project ./src/mcpTodo.ClientApp
   

Known Issues: Installing Foundry Local CLI

On Windows, if you keep failing to install Foundry Local CLI on your machine, try this way as a workaround:

# Download the package and its dependency
$releaseUri = "https://github.com/microsoft/Foundry-Local/releases/download/v0.3.9267/FoundryLocal-x64-0.3.9267.43123.msix"
Invoke-WebRequest -Method Get -Uri $releaseUri -OutFile ./FoundryLocal.msix
$crtUri = "https://aka.ms/Microsoft.VCLibs.x64.14.00.Desktop.appx"
Invoke-WebRequest -Method Get -Uri $crtUri -OutFile ./VcLibs.appx

# Install the Foundry Local package
Add-AppxPackage ./FoundryLocal.msix -DependencyPath ./VcLibs.appx

More resources

• Microsoft.Extensions.AI
• Anthropic SDK
• Google Vertex AI SDK
• Azure AI Foundry
• Foundry Local
• MCP SDK