MCP request successful but output not loaded into M365 Copilot

(1) Share

Report

Posted on by CU05110824-0

I create an agent named AgentX in COPILOT STUDIO. The AgentX consumes MS FABRIC DATA AGENT as MCP TOOLS. I have PUBLISHED the AgentX to MS TEAMS & M365 COPILOT (picture 1). In MS TEAMS, AgentX can run well. I tried using MS FABRIC DATA AGENT directly in M365 Copilot and it can run well. But when I try AgentX in M365 COPILOT, AgentX always returns the response "Sorry, I wasn't able to respond to that. Is there something else I can help with?" (picture 2). Even though when I look at the ACTIVITY TRANSCRIPT, the MCP TOOLS from MS FABRIC DATA AGENT can return a good response (picture 3). It can be seen from isError = false. Why does this happen? How to resolve this problem?

Categories:

Bot analytics

Model context protocol

Publish & channel management

I have the same question (0)

All responses (4)

Answers (0)

Sort by

Nivedipa-MSFT Microsoft Employee on at

Like (0)

Report
Copy link

Link copied!

Thanks for bringing this issue to our attention. We will check and update you soon.

Was this reply helpful? Yes No
Suggested answer

Valantis 3,406 on at

Like (1)

Report
Copy link

Link copied!
Hi @CU05110824-0 ,

Microsoft 365 Copilot is filtering or not rendering parts of AgentX’s final reply that work in Teams, so it shows “Sorry, I wasn’t able to respond to that,” even though the Fabric MCP tool returns isError = false.

try this fix

Return only supported content: plain text or a simple Adaptive Card (no Action.Execute). Don’t send Basic cards, images, video, files, or speech; avoid HTML/embedded links.

Normalize Fabric output (JSON/tables) into concise plain text or a minimal static card.

Publish, then start a new @AgentX chat. If needed, toggle off/on the “Teams and Microsoft 365 Copilot” channel and republish; optionally disable/re‑enable the app in Teams admin center.

Verify required services aren’t blocked and there are no DLP violations; keep responses compact.

Best regards,

Valantis

✅ If this helped solve your issue, please Accept as Solution so others can find it quickly.

❤️ If it didn’t fully solve it but was still useful, please click “Yes” on “Was this reply helpful?” or leave a Like :).

🏷️ For follow-ups @Valantis.

📝 https://valantisond365.com/

Was this reply helpful? Yes No
CU05110824-0 56 on at

Like (1)

Report
Copy link

Link copied!

Hi @Valantis

I create an agent in copilot studio

The agent consume an external agent from fabric data agent as an MCP tool

I publish the copilot studio agent in M365 Copilot, but it always return "Sorry, I wasn't able to respond to that. Is there something else I can help with?"

I also publish the fabric data agent in M365 Copilot, but it works well

The copilot studio agent can only return an answer if it is an AI Generated Answer that does not consume fabric data agent as an MCP tool (picture 1)

Why does this happen?

Was this reply helpful? Yes No
Suggested answer

Valantis 3,406 on at

Like (1)

Report
Copy link

Link copied!

Hi @CU05110824-0,

The key difference between your scenarios is authentication. The Fabric Data Agent works directly in M365 Copilot because it authenticates as you. When AgentX calls it as an MCP tool, the agent needs to pass your credentials to the tool, and that only works if authentication is configured correctly.

check:

1. Go to AgentX Settings > Security > Authentication and make sure it is set to Authenticate with Microsoft. If it is set to No authentication the agent cannot use tools that require user credentials, which would explain exactly what you are seeing.

2. Open the Fabric MCP tool settings inside AgentX and check whether the credential type is set to End user credentials. That is the correct setting for M365 Copilot so the tool runs as the signed-in user.

After any changes republish the agent and start a fresh AgentX conversation in M365 Copilot.

If this does not fix it, can you share a screenshot of AgentX Settings > Security > Authentication so i can see what it is currently set to? That will help us confirm the root cause.

Best regards,

Valantis

✅ If this helped solve your issue, please Accept as Solution so others can find it quickly.

❤️ If it didn’t fully solve it but was still useful, please click “Yes” on “Was this reply helpful?” or leave a Like :).

🏷️ For follow-ups @Valantis.

📝 https://valantisond365.com/

Was this reply helpful? Yes No