FAI MCP C# Generator

Scaffold a production-ready C#/.NET MCP server.

Parameters

• Server Name: ${SERVER_NAME="MyMcpServer"}
• Tools: ${TOOL_COUNT="1|3|5|10"}
• Azure Integration: ${AZURE="None|KeyVault|Search|OpenAI|Full"}

Generated Files

{SERVER_NAME}/
├── {SERVER_NAME}.csproj       # NuGet refs: ModelContextProtocol, Azure.Identity
├── Program.cs                  # Host builder with MCP server setup
├── Tools/
│   └── ExampleTools.cs         # [McpServerTool] sample implementations
├── Services/
│   └── IExampleService.cs      # DI interface pattern
├── Dockerfile
├── README.md
└── appsettings.json            # Non-secret configuration

Key Patterns

// Program.cs
var builder = Host.CreateApplicationBuilder(args);
builder.Services
    .AddMcpServer()
    .WithStdioTransport()
    .WithTools<ExampleTools>();
builder.Services.AddSingleton<IExampleService, ExampleService>();
var host = builder.Build();
await host.RunAsync();

Verification

• dotnet build compiles without errors
• dotnet run starts MCP server on stdio
• [Description] attributes present on all tools and parameters
• DefaultAzureCredential used for Azure services (no keys)

Tool Implementation Pattern

[McpServerTool, Description("Search FAI solution plays by keyword")]
public static async Task<string> SearchPlays(
    [Description("Search query")] string query,
    [Description("Max results (1-20)")] int maxResults = 5,
    IPlaySearchService searchService = null!)
{
    ArgumentException.ThrowIfNullOrWhiteSpace(query);
    maxResults = Math.Clamp(maxResults, 1, 20);
    var results = await searchService.SearchAsync(query, maxResults);
    return JsonSerializer.Serialize(results, new JsonSerializerOptions 
    { WriteIndented = true, PropertyNamingPolicy = JsonNamingPolicy.CamelCase });
}

Azure Integration Patterns

Key Vault

builder.Configuration.AddAzureKeyVault(
    new Uri("https://my-vault.vault.azure.net/"),
    new DefaultAzureCredential());

Azure OpenAI

builder.Services.AddSingleton(_ => new AzureOpenAIClient(
    new Uri(config["AzureOpenAI:Endpoint"]!),
    new DefaultAzureCredential()));

AI Search

builder.Services.AddSingleton(_ => new SearchClient(
    new Uri(config["Search:Endpoint"]!),
    config["Search:IndexName"],
    new DefaultAzureCredential()));

Testing

[Fact]
public async Task SearchPlays_ReturnsResults()
{
    var mockService = new Mock<IPlaySearchService>();
    mockService.Setup(s => s.SearchAsync("rag", 5))
        .ReturnsAsync(new[] { new PlayResult("01", "Enterprise RAG") });
    var result = await ExampleTools.SearchPlays("rag", 5, mockService.Object);
    Assert.Contains("Enterprise RAG", result);
}

Deployment

NuGet Dependencies

Configuration Reference

{
  "McpServer": { "Name": "my-fai-mcp", "Version": "1.0.0", "Transport": "stdio" },
  "AzureOpenAI": { "Endpoint": "https://my-oai.openai.azure.com/", "DeploymentName": "gpt-4o" },
  "Search": { "Endpoint": "https://my-search.search.windows.net", "IndexName": "fai-knowledge" }
}