openai -agent-dotnet

This MCP server is a .NET-based agent for interacting with an MCP-enabled TypeScript server that runs the to-do workflow. The agent communicates with the remote MCP server to manage AI-assisted to-do tasks via Azure OpenAI, while the client UI (Blazor) handles authentication and user interaction. To use, ensure the MCP client and server are deployed in a compatible environment (e.g., Azure Container Apps) and that the client is configured to talk to the server. The agent supports adjusting model selection (default gpt-5-mini) and authentication modes. Typical usage involves starting the server locally or deploying to your environment, then using the client app to issue prompts like listing to-dos, adding items, or marking tasks complete; the MCP server handles orchestration with the TypeScript server backend and Azure OpenAI.

Prerequisites:

• .NET 9 SDK
• Visual Studio Code with C# Dev Kit (optional but recommended)
• Node.js LTS (for any frontend tooling)
• Docker Desktop or Podman Desktop
• Azure CLI (az) and an Azure subscription

1. Clone the repository or download the sample package.

2. Restore and build the .NET agent project:

# replace with actual project path if different
cd path/to/openai-mcp-agent-dotnet

# restore dependencies
dotnet restore

# build the project
dotnet build -c Release

3. Run the agent locally (assuming the agent dll is built):

dotnet openai-mcp-agent-dotnet.dll

4. Optional: containerize or deploy to Azure following the guidance in the readme, using azd or your preferred deployment method. If deploying via azd, ensure environment variables for login, model name, and authentication are set as described in the prerequisites and notes.

5. Verify connectivity:

• Ensure the MCP client UI is deployed and can authenticate with ACA or your chosen auth method.
• Confirm the client and server communicate with the remote MCP server (TypeScript) and that Azure OpenAI is accessible with the configured model.

Notes and tips:

• By default, the MCP client is protected by ACA built-in authentication; you can disable this by setting USE_LOGIN to false during deployment.
• The agent uses gpt-5-mini by default; you can override with GPT_MODEL_NAME in your environment.
• If you enable API key authentication for Azure OpenAI (AOAI_USE_API_KEY true), ensure you provide the necessary credentials in your environment.
• Development mode settings (ENABLE_DEVELOPMENT_MODE) can help surface more verbose logs for troubleshooting.
• When deploying to Azure, the MCP server and client should be in the same resource group or properly networked to allow secure communication.
• Review the remote MCP server (TypeScript) docs for any endpoint access requirements and model deployment details.