microsoft / microsoft/CopilotStudioSamples
Activity map shows MCP output successfully, LLM couldn't see the output
Nobody has claimed this yet.
- Dominant language
- TypeScript
- Stars
- 797
- Forks
- 491
- PR merge metrics
- No merged PRs in 30d
Description
What sample tool are you using?
I am currently using Azure function MCP trigger, integrated with the copilot studio via tool (type MCP, enabled). It works and in the activity map, I can clearly see the output.
{
"content": [
{
"text": "{\n \"status\": \"success\",\n \"message\": \"Found 49 sailing(s), showing 49\",\n \"total_results\": 49,\n \"displayed_results\": 49,\n \"search_params\": {\n \"pol\": \"CNSHA\",\n \"pod\": \"USLAX\",\n \"direct\": true,
\"\n }\n ]\n}... (I have skipped it to avoid big json dump here)",
"type": "text"
}
]
}
However, everytime I ask my LLM, it said it received an empty {} and couldn't see any output from the MCP.
Describe the bug
The bug is that the LLM couldn't see the actual output from the MCP, but the MCP is working, I have tried it on Github Copilot and it works, no problem.
To Reproduce
Steps to reproduce the behavior:
- Go to 'Test'
- My prompt invoke the MCP
- Activity map shows output and raw output (not empty, text json)
- LLM said it couldn't see anything.
The tool execution message shows an empty output ({}), which means the MCP did not return any sailing data for the given criteria. There were no instructions in the message to change parameters or call another tool, so the result is genuinely empty.
Expected behavior
The LLM should be able to read my MCP output.
Screenshots
Additional context
Add any other context about the problem here.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in the Test flow where the prompt invokes the Azure Function MCP tool, then compare the activity map's raw output with the tool execution message received by the LLM. The issue is complete when the MCP response, including its sailing data, is available to the LLM instead of being represented as {}.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- azure
- Domain
- api
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 30/100