Feature: Show model name in tool call UI for subagents/Task tool
The Problem
When Claude Code spawns subagents via the Task tool, there's zero visibility into which model each subagent is using. The tool call UI header (e.g., Read, Bash, Searched) only shows the parent conversation's model — not the subagent's.
This matters a lot when you're running mixed-model workflows! For example:
- Parent on Opus 4.6 orchestrating architecture
- Explore agents inheriting Opus for deep research
- Sonnet agents handling boilerplate
Right now it all looks the same in the UI. You can't tell if your expensive Opus agent is doing the work or if it delegated to Sonnet. This makes it really hard to:
- Verify model selection — confirm subagents are using the model you specified (or inheriting correctly)
- Debug performance — understand if slowness is from model choice or task complexity
- Optimize cost — see where Opus is being used vs where Sonnet would suffice
- Build trust — know exactly what's running under the hood
What I'd Love to See
When a Task subagent is executing, show the resolved model name in the tool call header alongside the agent type. Something like:
⏺ Explore (Opus 4.6) | Read src/components/Button.tsx
⏺ general-purpose (Sonnet 4.5) | Bash npm test
⏺ Plan (Opus 4.6) | Grep "authentication"
Or even just a subtle label on the Task tool call itself:
⏺ Task (Explore → Opus 4.6) | "Find auth patterns"
Current Workaround
I built a PreToolUse/PostToolUse hook on the Task tool that writes the subagent model to a cache file, then reads it from the statusline script. It works, but it's a hack — the statusline updates on a poll interval, misses fast agents, and can't show per-tool-call model info the way the built-in UI can.
Why This Would Be Amazing
Claude Code's multi-agent orchestration is genuinely powerful. Being able to see the model flowing through each agent would make the experience so much more transparent and trustworthy. It's the kind of small UI detail that turns a black box into a glass box.
Would absolutely love to see this ship! 🚀
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗