Azure AI Foundry Agent Published to Teams/M365 Copilot Now Shows Code Interpreter Tool Calls to Users

Question

I have an Azure AI Foundry agent that is published to Microsoft Teams / Microsoft 365 Copilot.

Previously, when the agent used Code Interpreter (or other tools), end users only saw the final response generated by the agent.

Recently, I observed a change in behavior:

• The agent still functions correctly.
• However, users can now see tool execution details in the Teams/M365 Copilot chat experience.
• Code Interpreter activity and tool invocation information are being displayed to users.
• This behavior was not present previously for the same agent.

No major changes were made to the agent instructions or implementation prior to observing this behavior.

Questions

1. Has there been a recent change in how Azure AI Foundry agents are rendered in Microsoft Teams or Microsoft 365 Copilot?
2. Is displaying Code Interpreter/tool execution details now expected behavior?
3. Is there a setting to suppress tool invocation details and show only the final response?
4. Are there differences in behavior between:
   • Azure AI Foundry Playground
     • Microsoft Teams
       • Microsoft 365 Copilot
       1. Is this a known issue or regression?

Additional Details

• Agent created in Azure AI Foundry.
• Published using the Teams/M365 Copilot publishing workflow.
• Code Interpreter tool enabled.
• Previously only final responses were visible.
• Now tool execution details are visible in the chat experience.

Any clarification on expected behavior or available configuration options would be appreciated.ated.

Answer 1

Hello @Debesh Mohanty ,

Welcome to Microsoft Q&A .Thank you for reaching out to us.

The observed behavior appears to be associated with how agent activity is displayed within Teams and Microsoft 365 Copilot, but the underlying cause has not been publicly documented.

Regarding displaying Code Interpreter or tool execution details expected behavior - Code Interpreter operates as part of the agent execution process and can generate intermediate actions before producing a final response. In this scenario, those intermediate activities are being surfaced to end users within the chat experience. While the behavior is technically valid, it is currently unclear whether this presentation is intended for all published agents or whether the experience has changed over time.

As asked if there is a setting to suppress tool invocation details and show only the final response - At present, no known configuration has been identified that specifically controls the visibility of tool-execution activity within Teams or Microsoft 365 Copilot for published Foundry agents. The practical focus therefore needs to be on validating how the host experience is rendering agent activity rather than adjusting agent execution settings.

Please check if the following steps help-

1. Compare the same prompt across Azure AI Foundry Playground, Microsoft Teams, and Microsoft 365 Copilot.
2. Validate whether the behavior is consistent across multiple users and conversations.
3. Review agent instructions to ensure responses are optimized for concise final-answer generation.
4. Identify when the behavior was first observed and whether it coincided with any service-side changes in the environment.

The following are the differences between Azure AI Foundry Playground, Teams, and Microsoft 365 Copilot-

1. Azure AI Foundry Playground - Primarily intended for development, testing, and validation scenarios.
2. Microsoft Teams and Microsoft 365 Copilot - Provide separate end-user experiences and may present agent outputs differently from the Playground. Because of these differences, identical agent executions may appear differently across environments even when the underlying agent behavior remains unchanged.

The following references might be helpful , please check them out

• Publish agents to Microsoft 365 Copilot and Microsoft Teams - Microsoft Foundry | Microsoft Learn
• Use Code Interpreter with Microsoft Foundry agents - Microsoft Foundry | Microsoft Learn
• Agent tools overview for Microsoft Foundry Agent Service - Microsoft Foundry | Microsoft Learn
• Code Interpreter Capability for Declarative Agents for Microsoft 365 Copilot | Microsoft Learn  

Thank you

 

Please "Accept" the answer with an "Upvote" if the response was helpful. This will be benefitting other community members who face the same issue.

Answer 2

The provided information does not describe or document any change in how Azure AI Foundry agents are rendered in Microsoft Teams or Microsoft 365 Copilot, nor does it mention UI behavior around exposing Code Interpreter or other tool invocations to end users.

The context only covers:

• What Code Interpreter is and how to enable it in agents.
• How to use Code Interpreter with Foundry agents (Python, REST, Java, C#, declarative agents, Copilot Studio, etc.).
• Troubleshooting Code Interpreter execution, file handling, and timeouts.
• Tool availability across classic and new Foundry agents.

From this context, the following points are supported:

1. Code Interpreter is a standard tool for agents (both in Microsoft 365 Copilot declarative agents and Azure AI Foundry agents) and is expected to run in a sandboxed environment for data analysis, math, charts, and file processing.
2. Code Interpreter availability and behavior depend on the underlying provider and host; support and behavior can vary by host experience.
3. There is no documented setting here that controls whether tool invocation details (such as Code Interpreter steps) are visible or hidden in the Teams or Microsoft 365 Copilot UI.
4. There is no documentation in this context about:
   • A recent change in Teams/M365 Copilot rendering of Foundry agents.
   • A configuration option to suppress tool execution details and show only the final response.
   • Differences in UI behavior between Foundry Playground vs. Teams vs. Microsoft 365 Copilot for tool visibility.
   • Any known regression or issue related to tool call visibility.

Based on the available documentation, only these actions are clearly supported:

• Ensure Code Interpreter is enabled on the agent and supported by the chosen model and region.
• Use the troubleshooting guidance for execution issues (timeouts, missing files, corrupt outputs, unsupported packages, etc.).
• Use the appropriate SDKs/clients (C#, Python, Java, REST) to work with Code Interpreter and inspect code/output programmatically.

No additional guidance, flags, or UI configuration options related to hiding or changing the visibility of tool invocations in Teams or Microsoft 365 Copilot is present in the provided material.

---

References:

• Code Interpreter tool for Microsoft Foundry agents (csharp)
• Code Interpreter tool for Microsoft Foundry agents (rest)
• Code Interpreter tool for Microsoft Foundry agents (python)
• Migrate to the new agents developer experience
• Code Interpreter tool for Microsoft Foundry agents (java)
• Add the code interpreter capability to your agent
• Use code interpreter on SharePoint sources in agent conversations
• Code Interpreter (programming-language-csharp)