Model providers

Available providers

Any inference service that provides a Microsoft.Extensions.AI.IChatClient implementation can back a ChatClientAgent.

• Azure OpenAI — Azure-hosted OpenAI inference with Azure identity support.
• OpenAI — OpenAI Chat Completions and Responses APIs.
• Microsoft Foundry — Model inference through a Microsoft Foundry project.
• Anthropic — Claude model inference through Anthropic and supported hosted endpoints.
• Ollama — Local open-source model inference.
• Amazon Bedrock — AWS-managed foundation model inference.
• Google Gemini — Gemini Developer API or Vertex AI inference.
• ONNX — Local ONNX Runtime GenAI inference.
• Dapr — Inference routed through the Dapr Conversation building block.

Conversation history support

The selected API determines whether the remote service can own conversation history and whether the agent can instead use an in-memory or custom ChatHistoryProvider.

Service-managed history availability can also depend on the selected service options. See the provider page for configuration details.

SDK and endpoint selection

Several .NET SDKs can connect to Microsoft Foundry, Azure OpenAI, OpenAI, or Anthropic. Choose the SDK that matches the service endpoint and authentication model.

Use the Microsoft Foundry, Azure OpenAI, OpenAI, or Anthropic page for client construction and authentication examples.

Available providers

Agent Framework Python exposes provider-specific chat clients behind the common agent interface.

• Azure OpenAI — Azure-hosted OpenAI inference with Azure identity support.
• OpenAI — OpenAI Chat Completions and Responses APIs.
• Microsoft Foundry — Model inference through a Microsoft Foundry project.
• Foundry Local — Run supported Foundry models locally.
• Anthropic — Claude inference through Anthropic, Foundry, Amazon Bedrock, or Vertex AI.
• Ollama — Local open-source model inference.
• Amazon Bedrock — AWS-managed foundation model inference.
• Google Gemini — Gemini Developer API or Vertex AI inference.
• Mistral — Mistral AI embedding generation.

Available providers

The Go SDK creates a standard *agent.Agent through provider-specific constructors.

For example, create a Foundry-backed agent from a project endpoint, credential, and model deployment:

a := foundryprovider.NewAgent(endpoint, token, foundryprovider.ModelDeployment(model), foundryprovider.AgentConfig{
    Instructions: "You are a helpful assistant.",
    Config: agent.Config{
        Name: "MyAgent",
    },
})