Regarding Calling Orchestrator Agent from Sub Agent

(0) Share

Report

Posted on by PG-13041328-0

I have a topic inside the Sub agent , which generates a response but if any error comes , then it should go back to Orchestrator agent's topic which is having a RAG(knowlegde base) attached with him as a tool which can be used to generate the answer and if not able to get answer from it then post a message in appropriate teams channel .

So the main issue i am facing is how to reach orchestrator agent from the sub agent , how to do that ??

Categories:

Topic creation & management

I have the same question (0)

All responses (3)

Answers (0)

Sort by

Suggested answer

Beyond The Platforms 145 on at

Like (0)

Report
Copy link

Link copied!
Hi there!

The issue is an architectural one — multi-level agent chaining isn't supported in Copilot Studio: sub-agents cannot have their own connected agents Microsoft Learn, so a sub-agent calling back to the orchestrator creates a circular dependency that the platform won't allow.

The fix: invert the logic. The orchestrator should own the error handling, not the sub-agent.

When you redirect to a child agent from a topic, once the agent is done the originating topic resumes — and agents support passing output variables back to the parent. Microsoft Learn Use this to your advantage:

In the sub-agent → set an output variable (e.g. ErrorOccurred = true) when it fails

In the orchestrator topic → after the sub-agent node, add a Condition: if ErrorOccurred = true, query the RAG knowledge base tool; if RAG also has no answer, trigger a Power Automate flow that posts to the Teams channel

The orchestrator manages high-level decisions, while sub-agents focus on execution Microsoft Learn — so RAG and Teams posting should be tools on the orchestrator, not on the sub-agent.

Full architecture reference: Orchestrator and subagent patterns – Microsoft Learn

Hope this helps!
Paolo

✅ Did this solve your issue? → Accept as Solution
👍 Partially helpful? → Click "Yes" on "Was this reply helpful?" or drop a Like!

Want more tips on Power Platform & AI? Follow me here:

🔗 LinkedIn: https://www.linkedin.com/in/paoloasnaghi/
▶️ YouTube: https://www.youtube.com/@BeyondThePlatforms
📸 Instagram: https://www.instagram.com/beyond_the_platforms/
🌐 Website: https://www.beyondtheplatforms.com/

Was this reply helpful? Yes No
Suggested answer

AP-26031104-0 Microsoft Employee on at

Like (0)

Report
Copy link

Link copied!
Thanks for raising this issue.

In the current Orchestrator–Sub-agent pattern, sub-agents are not designed to directly call or “reach back” to the orchestrator. The orchestrator remains the single control and routing point for the entire workflow.

Instead of sub-agent → orchestrator invocation, the correct design is to use structured response handling:

The sub-agent should return either a success response or a structured error response (e.g., status: error, error_type, context).

The orchestrator then evaluates the response and decides the next step in the flow.

In your case, the recommended flow would be:

Orchestrator invokes sub-agent topic

If sub-agent returns success → proceed with response

If sub-agent returns error → orchestrator triggers RAG tool

If RAG also fails / low confidence → orchestrator posts message to Teams channel

This approach ensures proper separation of concerns and avoids circular agent dependencies, which are not supported in this design.

So to answer the main question: there is no supported way for a sub-agent to directly reach the orchestrator by design — this must be handled at the orchestrator level via fallback logic.

Was this reply helpful? Yes No
AP-26031104-0 Microsoft Employee on at

Like (0)

Report
Copy link

Link copied!

PG-13041328-0 Just following up on the answer I shared recently. Could you please let me know if it resolved your issue, or if you need any further clarification?

Was this reply helpful? Yes No