Power Platform Community Forum Thread Details

Hi,

Our Copilot Studio agent has suddenly stopped working. It was functioning
perfectly before, but now returns errors on every query — including simple ones.

## Environment
- Copilot Studio agent deployed as Teams app
- Connector: Windsor MCP (connecting to Google Analytics 4)
- Region: Japan

## Error Details

**Error 1:**
- Error Code: ConnectorPowerFxError
- Message: Evaluating PowerFx expression resulted in error while invoking the connector

**Error 2:**
- Error Code: OpenAIModelTokenLimit
- Conversation Id: a:1rpXIYWxCV1wMkA8pJnoGk9MsAmmMKUGs4sze4T1C6qX2SU7njJNHKU_nI0vKHlrEQ0HbXPFgZGG_2I_Qa_8wrn2Oz2fngdU7pu8GbfE6RBKlY4yInqqTlyBH-MO565DZ
- Time (UTC): 2026-07-14T01:14:56.289Z

## Troubleshooting Already Done

1. Created a brand new separate test agent (identical setup)
→ Same errors occur
2. Test agent is used by only 1 person
→ Same errors occur
3. Simple analysis queries that previously worked
→ Now fail with the same errors
4. Complex analysis that previously worked
→ Also fails
5. External connector (Windsor MCP → GA4) verified working correctly
6. No changes made to agent configuration or setup

## Timeline
- Previously: Agent worked perfectly for all types of analysis
(including complex queries exceeding 500KB responses),
tested with 4 users via Teams app with no issues
- Currently: All queries fail regardless of complexity

## Suspected Cause
This appears to be a Microsoft platform-side issue.
We found similar reports referencing Issue Number 6066935
for OpenAIModelTokenLimit.

## Request
1. Is this a known ongoing issue?
2. Can Microsoft investigate using the Conversation ID above?
3. What is the expected timeline for resolution?

Thank you.

Categories:

Building Copilot Studio chatbots in Microsoft Teams

Model context protocol

Hi,

Thank you for the detailed information.
Based on your testing, this does not appear to be isolated to a specific agent configuration, user, or query. Since the issue is reproducible across multiple agents and all query types, the behavior is consistent with a potential platform-side service issue or a change in platform limits.
Regarding your questions:

Is this a known ongoing issue?
- We are aware that other customers have reported errors related to OpenAIModelTokenLimit. However, we cannot confirm whether your specific issue is associated with an active incident based solely on the information provided.
Can Microsoft investigate using the Conversation ID?
- Yes. The Conversation ID and UTC timestamp you provided are valuable diagnostic details and should be included when opening a Microsoft Support ticket. Microsoft support engineers can use this information to review backend telemetry.
What is the expected timeline for resolution?
- We do not have visibility into active Microsoft engineering investigations or estimated resolution timelines. If this is tied to a platform incident, updates would typically be communicated through Microsoft Support.

Additional observations

The simultaneous occurrence of OpenAIModelTokenLimit and ConnectorPowerFxError across multiple agents suggests the issue may not be specific to the Windsor MCP connector.
The fact that previously working simple queries now fail strengthens the case for a service-side regression rather than an agent configuration issue.
Since a newly created test agent exhibits the same behavior, recreating the agent is unlikely to resolve the issue.

We recommend opening a Microsoft Support case and providing:

Conversation ID
UTC timestamp
Agent name
Environment name
Region (Japan)
Screenshots of both error messages
Confirmation that the issue reproduces in a newly created test agent

This will allow Microsoft to review the underlying service telemetry and determine whether the issue is related to an active platform incident or a recent service change.

Thanks,

Manoj Annavajjala