Can Claude models be used with Foundry Agent Service's create_agent API (threads/runs pattern) for portal visibility?

Hello @Tamasa Patra

Thank you for your detailed question. I understand how important it is to have a consistent enterprise experience—including governance, observability, tracing, and portal-based lifecycle management—across all models used in your multi-agent architecture.

At this time, Claude models deployed through the Azure AI Foundry model catalog are not supported by Azure AI Foundry Agent Service's create_agent API (the threads/runs managed-agent pattern exposed through AIProjectClient.agents.create_agent).

As a result, you cannot currently create a Foundry-managed agent using a Claude deployment such as:

AIProjectClient.agents.create_agent(

and have that agent:

• Appear in the Agents tab in the Azure AI Foundry portal
• Support native thread and run management
• Provide built-in tracing and observability
• Benefit from the full Foundry Agent Service governance and lifecycle management experience

Why This Happens

Today, the Foundry Agent Service supports only:

• Azure OpenAI models
• Select models sold directly by Azure (including supported Azure Marketplace models) that are explicitly listed in the Agent Service model support documentation

Although Claude is available as a first-class model in the Azure AI Foundry model catalog for inference, it is not currently included in the list of models supported by Foundry Agent Service.

What This Means for Claude Agents

Claude-powered agents created using any of the following:

• Anthropic SDK via /anthropic/v1/messages
• Microsoft Agent Framework with AnthropicFoundryClient
• Claude Agent SDK

are treated as externally orchestrated applications. Therefore, they:

• Do not appear in the Foundry portal's Agents tab
• Do not support native Foundry Agent Service threads/runs
• Do not automatically inherit built-in tracing, run history, or portal-based management

Semantic Kernel Support

Similarly, Semantic Kernel's AzureAIAgent class relies on the same Foundry Agent Service backend. Because Claude is not currently supported by that service, AzureAIAgent cannot presently use a Claude deployment as its underlying model.

Clarification on Documentation

When documentation states that Claude models are available with "Foundry APIs, SDKs, agents, and governance features," the term "agents" refers broadly to:

• Building agentic solutions using Claude within Azure AI Foundry
• Using orchestration frameworks such as Microsoft Agent Framework or custom implementations
• Applying Azure governance, security, and compliance controls at the platform level

It does not currently refer to support for the Foundry Agent Service create_agent managed-agent framework.

Current Recommendation

If your requirements include:

• Portal visibility in the Foundry Agents tab
• Native thread/run management
• Built-in tracing and observability
• Full Foundry Agent Service governance controls

then the recommended approach today is to use a supported:

• Azure OpenAI model, or
• Eligible Azure Marketplace model supported by Agent Service

If Claude-specific capabilities are essential, you will need to continue using one of the external orchestration approaches listed above until native Agent Service support becomes available.

Support for additional models in Foundry Agent Service continues to evolve. However, there is currently no publicly announced timeline for Claude support within the create_agent / threads-runs Agent Service framework.

We encourage you to submit or upvote a feature request through the Azure feedback channels, as customer demand plays an important role in prioritization.

Please refer this

Models supported by Foundry Agent Service https://learn.microsoft.com/azure/ai-foundry/agents/concepts/model-region-support

I Hope this helps. Do let me know if you have any further queries.

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thank you!

1. Claude models and Foundry Agent Service create_agent

The available documentation shows:

• Claude models are supported in Microsoft Foundry for direct model inference via the Anthropic-compatible /anthropic endpoint and via Foundry SDKs and REST APIs.
• Foundry Agent Service supports “many models from the Foundry model catalog, such as GPT‑4o, Llama, and DeepSeek,” and exposes two patterns: Responses Agents (ChatClientAgent) and Foundry Agents (FoundryAgent) that are versioned and managed in the portal.
• The “Models supported by Agent Service” list referenced in the question is not present in the provided context, and there is no explicit statement that Claude models are supported as Foundry Agents via AIProjectClient.agents.create_agent.

Because the context does not state that Claude can be used as a model for Foundry Agents created via AIProjectClient.agents.create_agent, there is no basis to confirm that a Claude-powered agent created this way will appear in the Foundry portal Agents tab with full threads/runs integration.

2. Roadmap / future support

The context does not contain any roadmap or forward-looking statement about adding Claude to the “Models supported by Agent Service” list or enabling Claude as a first-class Foundry Agent Service model. No roadmap details can be provided from the given information.

3. Semantic Kernel AzureAIAgent with Claude

The context describes:

• Direct Anthropic usage via AnthropicClient and AnthropicFoundryClient in Microsoft Agent Framework, including environment variables for Foundry-based Anthropic usage.
• Foundry integration in Agent Framework via Microsoft.Agents.AI.Foundry, which supports both direct model inference and Foundry Agents.

However, there is no mention of Semantic Kernel’s AzureAIAgent class or its compatibility matrix with Claude deployments. The context does not state that AzureAIAgent can use Claude as the model parameter, so this cannot be confirmed.

4. What is supported today (from the provided context)

From the context, the supported Claude usage patterns in Foundry are:

• Direct deployment and inference of Claude models in Microsoft Foundry, authenticated via Entra ID or API keys, using the Anthropic Messages API (/anthropic/v1/messages) and Foundry SDKs/REST.
• Building agents with Claude via:
  • Microsoft Agent Framework Anthropic provider (AnthropicClient) using Anthropic public API.
  • Microsoft Agent Framework Anthropic Foundry provider (AnthropicFoundryClient) using Foundry-hosted Claude deployments.
• These agents are standard AIAgent instances in Agent Framework and benefit from framework-level orchestration, but the context does not state that they become Foundry Agent Service resources visible in the Foundry portal Agents tab.

The context also confirms that Claude models in Foundry:

• Require Enterprise or MCA‑E subscriptions.
• Have specific quotas and limits.
• Must be used with appropriate responsible AI and content safety configuration.

Given the available information, Claude is clearly a first-class model family in Foundry for inference and for use with Microsoft Agent Framework, but there is no explicit documentation here that Foundry Agent Service’s create_agent (threads/runs pattern with portal-managed agents) currently supports Claude as a model.

References:

• Deploy and use Claude models in Microsoft Foundry (preview)
• Anthropic Agents (programming-language-python)
• Anthropic Agents (programming-language-csharp)
• What is Microsoft Foundry Agent Service?
• Microsoft Foundry (programming-language-csharp)
• Are Anthropic models deployed via Azure AI Foundry officially supported with third-party tools such as HAI or Cline? - Microsoft Q&A