Answered

"explanation of tool call" message

(5) Share

Report

Posted on by TS-16120959-0

When conversing with an Agent linked to a Child Agent, the following message is output after the Child Agent provides a response. (This also occasionally occurs when conversing with an Agent not linked to a Child Agent.)

{"explanation_of_tool_call":"The user's query has been fully addressed ~~"}

I suspect this occurs when Copilot Studio's “internal metadata (explanation of inference)” is returned instead of the text for the user. How can I prevent this message from appearing?

Categories:

Calling actions from Copilot Studio

Autonomous agents

I have the same question (11)

All responses (4)

Answers (2)

Sort by

Suggested answer

Michael E. Gernaey 53,968 Super User 2026 Season 2 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi @TS-16120959-0

Can you share if you are simply displaying whatever the Agent Tool (whether its another agent, or power automate or agent flow) is sending you?

If so, you will need to review it, and strip out what you do not want.

If you can share the output from the XXXXX Tool, when it did this, we can review the JSON and show you how to strip out the metadata.

If these suggestions help resolve your issue, Please consider Marking the answer as such and also maybe a like.

Thank you!
Sincerely, Michael Gernaey

Was this reply helpful? Yes No
Verified answer

-jake- 38 on at

Like
a
(2)

Report
Copy link

Link copied!

I believe it's a known bug. After experiencing the same problem, I found an active discussion on github the other night. Someone from MSFT shared a defect or incident number in the thread, related to this exact problem. Seems like there's more than a few agents affected.

Good luck. Hopefully it will be resolved soon.

1 people found this reply helpful.

Was this reply helpful? Yes No
Verified answer

JV-20091504-0 13 on at

Like
a
(1)

Report
Copy link

Link copied!

I think it has to be a bug. I'm seeing the same with MCP tool call metadata being displayed in the response. In the short term, I've disabled generative orchestration and am filtering the junk out of the generative answers response.

1 people found this reply helpful.

Was this reply helpful? Yes No
Suggested answer

AV-11011923-0 2 on at

Like
a
(0)

Report
Copy link

Link copied!

Yes, it's a bug for sure as I am using same Adaptive card extracting AnswerText from AnswerRecord correctly for so many months. But when I redesigned and used multi agents to implement same design. I started getting this additional json eveytime and it's kind of funny that sometimes it will not and then will start coming every time. CoPliot suggested to use Activity.text to pass user query in place of LastMessage.text and when you switch first few calls will be without it and then again it will start coming. there is no switch to turn it off. if you use same Activity.text or LastMessage in same Agent with multiple topics it will not come. Here is my output of json.

{"explanation_of_tool_call":"The user’s request has been fully addressed by providing detailed steps on how to book a desk using the app, along with a source link for reference.","":""}.

seems not much testing has done on this and there is no way you can pass user Input from one agent to other external agent as well yet and hence dependent on these system level variables.

Was this reply helpful? Yes No