Fix subagent tool calling loops silently switching models #2932
+116
−15
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Subagents were silently falling back to GPT-4.1 when
endpoint.supportsToolCallsreturned false, causing tools to be unavailable on the user's selected model (e.g., Gemini 3 Pro). This resulted in sporadic failures with "Server error. Stream terminated" as models couldn't access required tools.Changes
Removed model switching logic from three subagent tool calling loops:
SearchSubagentToolCallingLoopCodebaseToolCallingLoopMcpToolCallingLoopAdded unit tests verifying subagents preserve the parent request's model regardless of
supportsToolCallsvalueBefore
After
Subagents now consistently use the same model as their parent request. If a model genuinely lacks tool support, it will fail explicitly rather than silently switching models.
Original prompt
💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.