Hello Power Platform Community,
We’ve built a Copilot Studio agent that uses two Dataverse tools:
- Search Tool – queries a Dataverse table based on the user’s input.
- Log Tool – inserts a record into another Dataverse table to store the search details (query text, timestamp, etc.).
The agent and tools are configured so that:
- The agent is instructed to invoke the Search Tool first, then the Log Tool.
- The Search Tool is also instructed to invoke the Log Tool immediately after completion.
- The Log Tool has instructions to run after the Search Tool finishes.
However, the behavior is inconsistent:
- Sometimes both tools run correctly — the Search Tool returns results, and the Log Tool logs the query.
- Other times, the Log Tool is not triggered at all.
- Occasionally, the Search Tool fails to return results.
- In some runs, the Log Tool incorrectly prompts for inputs that should be automatically passed from the Search Tool.
- It can even alternate between working and not working (e.g., runs fine on the first query, fails on the second, then works again).
Has anyone experienced similar reliability issues when chaining multiple Dataverse tools in an agent?
Is there a known limitation or timing issue with tool-to-tool invocation in Copilot Studio?
Any suggestions, best practices, or troubleshooting steps would be greatly appreciated!
