elbruno.Extensions.AI.Claude

elbruno.Extensions.AI.Claude

A C# library for using Claude models deployed in Microsoft Foundry, implementing the Microsoft.Extensions.AI abstractions.

Overview

This library provides an AzureClaudeClient that allows you to interact with Claude models (e.g., Claude Sonnet 4.5) deployed in Microsoft Foundry. It follows the same patterns as AzureOpenAIClient and implements the standard IChatClient interface from Microsoft.Extensions.AI.

Video Tutorial

Watch this 5-minute video to learn how to easily use this library in a set of sample projects:

Features

• ✅ Full IChatClient implementation
• ✅ Support for both streaming and non-streaming completions
• ✅ Azure authentication using Azure Identity or API keys
• ✅ System messages support
• ✅ Multi-turn conversations
• ✅ Token usage tracking
• ✅ Configurable chat options (temperature, top-p, max tokens)

Installation

dotnet add package elbruno.Extensions.AI.Claude

Prerequisites

1. An Azure subscription
2. A Claude model deployed in Microsoft Foundry
3. Appropriate Azure credentials configured

Quick Start

Authenticate with DefaultAzureCredential (recommended)

using Azure.Identity;
using elbruno.Extensions.AI.Claude;
using Microsoft.Extensions.AI;

var endpoint = new Uri("https://your-endpoint.services.ai.azure.com");
var deploymentName = "claude-sonnet-4-5";
var credential = new DefaultAzureCredential();

var client = new AzureClaudeClient(endpoint, deploymentName, credential);

var messages = new List<ChatMessage>
{
    new(ChatRole.User, "What is the capital of France?")
};

var response = await client.CompleteAsync(messages);
Console.WriteLine(response.Message.Text);

Authenticate with an API key (when managed identity isn't available)

using elbruno.Extensions.AI.Claude;
using Microsoft.Extensions.AI;

var endpoint = new Uri("https://your-endpoint.services.ai.azure.com");
var deploymentName = "claude-sonnet-4-5";
var apiKey = Environment.GetEnvironmentVariable("AZURE_CLAUDE_APIKEY")
    ?? throw new InvalidOperationException("Store the API key securely (for example in Azure Key Vault) and expose it via configuration.");

var client = new AzureClaudeClient(endpoint, deploymentName, apiKey);

var response = await client.CompleteAsync(new List<ChatMessage>
{
    new(ChatRole.User, "Share two secure ways to store secrets in Azure.")
});

Console.WriteLine(response.Message.Text);

Usage Examples

Simple Completion

var messages = new List<ChatMessage>
{
    new(ChatRole.User, "Explain quantum computing in simple terms.")
};

var response = await client.CompleteAsync(messages);
Console.WriteLine(response.Message.Text);
Console.WriteLine($"Tokens used: {response.Usage?.TotalTokenCount}");

Streaming Completion

var messages = new List<ChatMessage>
{
    new(ChatRole.User, "Write a short story about a robot.")
};

await foreach (var update in client.CompleteStreamingAsync(messages))
{
    if (!string.IsNullOrEmpty(update.Text))
    {
        Console.Write(update.Text);
    }
}

Multi-turn Conversation

var conversation = new List<ChatMessage>
{
    new(ChatRole.System, "You are a helpful coding assistant."),
    new(ChatRole.User, "How do I create a list in Python?"),
};

var response1 = await client.CompleteAsync(conversation);
conversation.Add(response1.Message);

conversation.Add(new ChatMessage(ChatRole.User, "Can you show me an example?"));
var response2 = await client.CompleteAsync(conversation);

With Chat Options

var options = new ChatOptions
{
    Temperature = 0.7,
    TopP = 0.9,
    MaxOutputTokens = 2048
};

var messages = new List<ChatMessage>
{
    new(ChatRole.User, "Generate creative ideas for a tech startup.")
};

var response = await client.CompleteAsync(messages, options);

Configuration

Environment Variables

You can configure the client using environment variables:

export AZURE_CLAUDE_ENDPOINT="https://your-endpoint.services.ai.azure.com"
export AZURE_CLAUDE_MODEL="claude-sonnet-4-5"
export AZURE_CLAUDE_APIKEY="your-api-key" # Prefer Azure Key Vault or managed identity instead of plain env vars

Authentication

The library supports two authentication flows:

1. Azure Identity (recommended) – Works with DefaultAzureCredential, ManagedIdentityCredential, EnvironmentCredential, AzureCliCredential, etc. Use this whenever you can rely on managed identity, service principals, or developer credentials.
2. API Key authentication – Useful for constrained environments where Azure Identity is not available. Store API keys in Azure Key Vault or another secure secret store; never hardcode them.

Using Azure Identity

• DefaultAzureCredential (recommended for development and production)
• ManagedIdentityCredential (for Azure resources)
• EnvironmentCredential (for environment variables)
• AzureCliCredential (for local development)

Example with specific credential:

var credential = new ManagedIdentityCredential();
var client = new AzureClaudeClient(endpoint, deploymentName, credential);

Note on Azure AI Foundry headers

Claude deployments hosted in Azure AI Foundry require the anthropic-version: 2023-06-01 header for every request. Version 0.1.0-preview.3 of this library automatically applies it for both Managed Identity and API key scenarios. If you're pinned to an older version, add the header manually or upgrade to avoid invalid_request_error responses.

Using an API key

var apiKey = configuration["AZURE_CLAUDE_APIKEY"]
    ?? throw new InvalidOperationException("Configure AZURE_CLAUDE_APIKEY via Key Vault, Azure App Configuration, or user secrets.");

var client = new AzureClaudeClient(endpoint, deploymentName, apiKey);

Supported Models

This library supports Claude models deployed in Microsoft Foundry through global standard deployment. The following models are currently available:

Available Models

• Claude Sonnet 4.5 (deployment name: claude-sonnet-4-5) - Anthropic's most capable model for building real-world agents and handling complex, long-horizon tasks. Best for agentic workflows, computer use capabilities, and production deployments.

• Claude Haiku 4.5 (deployment name: claude-haiku-4-5) - Delivers near-frontier performance with optimal speed and cost. One of the best coding and agent models, ideal for free products and scaled sub-agents.

• Claude Opus 4.1 (deployment name: claude-opus-4-1) - Industry leader for coding. Delivers sustained performance on long-running tasks requiring focused effort and thousands of steps, significantly expanding what AI agents can solve.

All models support:

• 1 million token context window
• Extended thinking for enhanced reasoning
• Image and text input
• Code generation, analysis, and debugging (Sonnet 4.5 and Opus 4.1)

Choosing the Right Model

• Claude Sonnet 4.5: Best for balanced performance and capabilities, production workflows, complex reasoning, and agentic tasks
• Claude Haiku 4.5: Best for speed and cost optimization, high-volume processing, fast responses
• Claude Opus 4.1: Best for complex coding tasks, long-running agent workflows, and enterprise applications requiring sustained performance

API Reference

AzureClaudeClient

Constructor

public AzureClaudeClient(
    Uri endpoint,
    string modelId,
    TokenCredential credential,
    HttpClient? httpClient = null)

public AzureClaudeClient(
    Uri endpoint,
    string modelId,
    string apiKey,
    HttpClient? httpClient = null)

Properties

• Metadata - Returns ChatClientMetadata with provider name and model ID

Methods

• CompleteAsync - Sends a completion request and returns the full response
• CompleteStreamingAsync - Sends a streaming completion request
• GetService - Gets a service of the specified type
• Dispose - Disposes resources

Building the Project

# Clone the repository
git clone https://github.com/elbruno/elbruno-extensions-ai-claude.git
cd elbruno-extensions-ai-claude

# Build the solution
dotnet build

# Run tests
dotnet test

# Pack the NuGet package
dotnet pack src/elbruno.Extensions.AI.Claude/elbruno.Extensions.AI.Claude.csproj

Testing

The project includes unit tests for core functionality. Run them with:

dotnet test

Sample Applications

Three console applications ship in the samples folder:

1. elbruno.Extensions.AI.Claude.Samples – demonstrates Azure Default Credentials (managed identity flow).
2. elbruno.Extensions.AI.Claude.ApiKeySample – demonstrates API key authentication. Store AZURE_CLAUDE_APIKEY securely and load it through configuration/User Secrets.
3. elbruno.Extensions.AI.Claude.AgentSample – demonstrates how to use the client with Microsoft.Agents.AI to create an AI agent.

Run the Default Credential sample:

# Set environment variables
export AZURE_CLAUDE_ENDPOINT="your-endpoint"
export AZURE_CLAUDE_MODEL="claude-sonnet-4-5"

# Run the sample
dotnet run --project samples/elbruno.Extensions.AI.Claude.Samples

Run the API key sample (after setting AZURE_CLAUDE_APIKEY securely):

export AZURE_CLAUDE_ENDPOINT="your-endpoint"
export AZURE_CLAUDE_MODEL="claude-sonnet-4-5"
export AZURE_CLAUDE_APIKEY="your-api-key"

dotnet run --project samples/elbruno.Extensions.AI.Claude.ApiKeySample

Run the Agent sample:

export AZURE_CLAUDE_ENDPOINT="your-endpoint"
export AZURE_CLAUDE_MODEL="claude-sonnet-4-5"
export AZURE_CLAUDE_APIKEY="your-api-key"

dotnet run --project samples/elbruno.Extensions.AI.Claude.AgentSample

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

References

• Microsoft Foundry Documentation
• Deploy and use Claude models in Microsoft Foundry
• Microsoft.Extensions.AI Documentation
• Claude API Documentation
• Azure Identity Library
• Generative AI for Beginners - .NET

Support

For issues and questions:

• Open an issue on GitHub
• Contact: El Bruno

Changelog

0.1.0-preview.6

• Updated documentation with video tutorial

0.1.0-preview.4

• Implemented GetResponseAsync to support Microsoft.Agents.AI
• Added Agent sample project

0.1.0-preview.2

• Added API key authentication constructor and helper methods
• Added dedicated API key console sample and documentation
• Updated NuGet metadata and release notes

0.1.0-preview.3

• Fixed DefaultAzureCredential flow by always including the required anthropic-version header
• Documented the Claude header requirement in Azure AI Foundry

0.1.0 (Initial Release)

• Initial implementation of AzureClaudeClient
• Support for completions and streaming
• Azure authentication support
• Full IChatClient interface implementation