How to set the Reasoning Effort for Foundry Agents?

Hello Sammy Lastre,

Welcome to Microsoft Q&A.

Thank you for reaching out regarding the “Reasoning Effort” parameter in the Azure AI Foundry Agents API.

At this time, the Reasoning Effort setting which allows you to control the model’s reasoning depth (low, medium, or high) is only available for direct model inference endpoints, such as the Completions or Responses API, when using reasoning-capable models (e.g., o1, o3-mini).

In the Agents API, this parameter is not currently supported. Based on the available documentation, there isn’t an option to specify “Reasoning Effort” directly when configuring or invoking an agent. The current focus of the Agents API is on managing multi-turn conversations, tool integrations, and context orchestration, rather than low-level reasoning controls.

As for future support, the official documentation does not yet indicate any upcoming feature or timeline for adding Reasoning Effort control to the Agents API. However, this is a known area of interest, and the product team is aware of customer demand for more flexible reasoning control within the agent experience.

If your scenario requires adjusting the reasoning depth today, you can consider one of the following workarounds:

Use the Responses API directly with a reasoning-capable model and specify the desired reasoning_effort value.

Include explicit reasoning-style instructions in your agent’s system or developer message (e.g., “reason step-by-step in detail” or “respond concisely with minimal reasoning”) to simulate a similar effect.

We recommend please keeping an eye on the official Azure AI Foundry documentation for any future updates on this feature.

Also refer this How to use reasoning models with Azure AI Foundry Models - Python

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!