In this article, I will explain how to use a Microsoft Foundry model in Copilot Studio.

If you are new to Microsoft Foundry and want to learn how to deploy a model, please refer to my previous article: Deploy models and Build an Agent

Getting Model Details from Foundry:

First, we need to collect the model connection details from the Microsoft Foundry portal.

  • Sign into Foundry Portal
  • Under Build > Models, choose the model you deployed.
  • Go Details tab. Keep this screen open, as we will use the information from this page when connecting the model to Copilot Studio.

Create ‘Prompt’ in Copilot Studio:

Microsoft Foundry models can be used in Copilot Studio by configuring a Prompt. Prompts allow you to define instructions and behavior for how the AI model should respond.

  • Sign in to Copilot Studio
  • Navigate to Tools > Prompt
  • Create a new Prompt
Connect the Microsoft Foundry Model:
  • Inside the Prompt configuration screen:
    • In the Model dropdown, select Azure AI Foundry Models > Connect a new model
  • Click on Create a new model
  • Copy the required details from the Model Summary / Details screen in the Foundry portal (which we opened earlier). Paste those values into the connection screen.
  • Click Connect
  • Once the connection is successful, the model status will show Connected.
Test the Prompt:

After the model is connected:

  • Go ahead and provide Instructions and test the Prompt using the Foundry model.
  • Copilot Studio will now send the prompt to the Microsoft Foundry model, and the response will be returned directly inside the test interface.

Important Note

  • Only chat-enabled Microsoft Foundry models can be used in Copilot Studio Prompts.
  • While writing this article, I initially tried using the gpt-5.1-codex-mini model deployed in Foundry, but it was not chat-enabled, which resulted in an error when adding it to the prompt.
  • To proceed with the example, I used the DeepSeek-V3.1 model instead, as it supports chat capabilities and works correctly with Copilot Studio prompts.
  • Always ensure the model you select in Foundry supports chat interactions before connecting it to Copilot Studio.

🙂

Advertisements
Advertisements

Leave a comment