How to Integrate OpenAI Models via OpenRouter in C# Using the Microsoft Agent Framework

Watch Video  Download SourceCode 

Introduction

Building intelligent applications has become a core requirement for modern software. However, hardcoding your applications to a single AI provider's endpoint can lead to rigid architectures and vendor lock-in.

Having spent over a decade architecting and building robust .NET applications, I’ve learned that flexibility is just as important as functionality. When building AI agents, you want the ability to hot-swap language models (like moving from GPT-4o to Claude 3.5 or Llama) without rewriting your entire integration layer.

This is where OpenRouter shines. By combining OpenRouter with the new Microsoft Agent Framework, we can create powerful, model-agnostic AI agents in C#. In this tutorial, I'll walk you through a clean, practical implementation of exactly how to achieve this.

Why OpenRouter and Microsoft Agent Framework?

• Microsoft Agent Framework (Microsoft.Agents.AI): Provides a standardized, elegant abstraction for defining AI agents, their personas, and their capabilities within the .NET ecosystem.

• OpenRouter: Acts as a unified API gateway. You use the standard OpenAI SDK format, but point it at OpenRouter to instantly unlock access to dozens of different LLMs (OpenAI, Anthropic, Meta, Google, etc.).

Account & Credentials

1. OpenRouter Account — Sign up at https://openrouter.ai/
2. OpenRouter API Key — Found in your account settings

Step 1: Setting Up the Configuration

First, we need to securely store our OpenRouter API key. Avoid hardcoding credentials directly into your C# files. Instead, add your API key to your appsettings.json file.

{
  "Logging": {
    "LogLevel": {
      "Default": "Information",
      "Microsoft.AspNetCore": "Warning"
    }
  },
  "AllowedHosts": "*",
  "OpenRouter": {
    "ApiKey": "YOUR_OPENROUTER_API_KEY_HERE"
  }
}

Step 2: Create Your .NET 10 Project

Let's start by creating a new ASP.NET Core Web API project. Open your terminal and execute the following command:

dotnet new webapi -n Agenframework -f net10.0
cd Agenframework

This command creates a new .NET 10 Web API project named "Agenframework".

Step 3: Install Required NuGet Packages

Navigate to your project directory and install the necessary NuGet packages:

dotnet add package OpenAI
dotnet add package Microsoft.Extensions.AI
dotnet add package Microsoft.Agents.AI
dotnet add package Swashbuckle.AspNetCore

Package Overview

Step 4: Create the Chat Controller with OpenRouter Integration

Create a new file called ChatController.cs in the Controllers folder:

using Microsoft.AspNetCore.Mvc;
using Microsoft.Extensions.AI;
using Microsoft.Agents.AI;
using OpenAI;
using OpenAI.Chat;
using System.ClientModel;

namespace Agenframework.Controllers
{
    [Route("api/[controller]")]
    [ApiController]
    public class ChatController : ControllerBase
    {
        IConfiguration _configuration;
        public ChatController(IConfiguration configuration)
        {
            _configuration = configuration;
        }

        [HttpPost("Chat")]
        public async Task<string> ChatAsync([FromBody] string UserInput)
        {
            string key = _configuration["OpenRouter:ApiKey"] ?? string.Empty;

            IChatClient chatClient =
                new ChatClient(
                    "openai/gpt-4o-mini",
                    new ApiKeyCredential(key),
                    new OpenAIClientOptions
                    {
                        Endpoint = new Uri("https://openrouter.ai/api/v1/")
                    })
                .AsIChatClient();

            AIAgent writer = new ChatClientAgent(
                chatClient,
                new ChatClientAgentOptions
                {
                    Name = "Writer",
                    Description = "Write stories that are engaging and creative."
                });

            var response = await writer.RunAsync(UserInput ?? "Write a short story about a haunted house.");
            return response.ToString();
        }
    }

}

Code Explanation

• Endpoint Configuration: The https://openrouter.ai/api/v1/ endpoint tells the OpenAI client to route requests through OpenRouter instead of OpenAI's direct endpoint.
• Model Selection: "openai/gpt-4o-mini" specifies that we want to use OpenAI's GPT-4o-mini model through OpenRouter.
• API Key Management: We're using configuration value instead of hardcoding the key for better security.
• AI Agent: The ChatClientAgent wraps the chat client with agent capabilities, allowing for more sophisticated interactions.

Step 6: Test Your API

Using Swagger UI

1. Run your application: dotnet run
2. Navigate to https://localhost:5001/swagger/index.html
3. Find the POST /api/chat/Chat endpoint
4. Click "Try it out"
5. Enter a prompt like: "Write a funny poem about programming"
6. Click "Execute"

Popular Model Options

• "openai/gpt-4o" — Most capable OpenAI model
• "openai/gpt-4-turbo" — Faster, more affordable GPT-4
• "openai/gpt-4o-mini" — Lightweight and cost-effective
• "anthropic/claude-3-opus" — Anthropic's most capable model
• "anthropic/claude-3-haiku" — Fast and lightweight Claude model
• "google/gemini-pro" — Google's Gemini model

FAQ: OpenRouter Integration with .NET 10

Q: What is the difference between using OpenRouter and calling OpenAI directly?

A: OpenRouter provides a unified interface for multiple providers, making it easier to switch models and manage costs. It also offers fallback routing if your preferred model is unavailable.

Q: Can I use different models for different requests?

A: Yes! As shown in the advanced configuration section, you can accept the model name as a parameter in your request.

Q: Is OpenRouter more expensive than calling OpenAI directly?

A: Pricing is typically comparable. OpenRouter sometimes offers better rates due to competitive pricing between providers.

Q: How do I handle rate limiting?

A: Implement rate limiting middleware in your ASP.NET Core application, as shown in the security best practices section.

Q: Can I use OpenRouter in a production environment?

A: Absolutely! OpenRouter is production-ready and used by many organizations. Ensure you follow the security best practices outlined in this guide.

Q: What other AI models can I use besides ChatGPT?

A: OpenRouter supports models from Anthropic (Claude), Google (Gemini), Meta (Llama), and many others. Check https://openrouter.ai/ for the complete list.

Resources

• OpenRouter Documentation
• OpenAI .NET SDK
• Microsoft.Extensions.AI Documentation
• ASP.NET Core Documentation

Final Thoughts

If you have any questions or run into issues, feel free to leave a comment below. I'm here to help!

Happy coding, and enjoy building AI-powered applications!