Subagent model not shown in UI when multiple agents launch in parallel
Environment: Claude Code CLI on Linux (terminal UI).
What happened:
When the assistant launches a single background subagent via the Agent tool, the UI shows which model the subagent is running on. When it launches two subagents in parallel (two Agent tool calls in one message), the model indicator is not visible for the launched agents.
Why it matters:
I enforce a policy that subagents run on a cheaper model (opus) than the controller. With parallel launches I could not verify the model from the UI and had to stop both agents to check; the assistant then had to confirm the model by grepping the agents' JSONL transcripts ("model":"claude-opus-5"). The agents were in fact on the requested model — this is purely a UI visibility gap.
Expected:
Parallel agent launches display the per-agent model the same way a single launch does.
Repro:
- Ask Claude to launch one background subagent with an explicit model override — model is visible in the UI.
- Ask it to launch two subagents in parallel in the same turn — model is not visible.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗