Claude Code's default subagent status line already shows the task description and token usage — but not which model each subagent is running on. This small custom status line script adds the model to the line.
When your main session runs an expensive model like Fable 5, subagents inherit that model unless you (or Claude) say otherwise. That means a Fable session can quietly delegate work to another Fable agent — and burn through your token limits much faster than you expected.
With the model in the status line, you can see at a glance when a subagent is running on a pricier model than the task needs, and interrupt or redirect before it eats your usage. In the screenshot above, the subagent is confirmed to be running claude-haiku-4-5-20251001 — a cheap model for a cheap task.
Each subagent's status line will show:
local_agent · claude-haiku-4-5-20251001 · Echo, sleep, done message · ↓ 28.8k tokens
The model is the new piece of information; the type, description, and token count are also included so the line stays a full replacement for the default.
Requirements: jq must be installed (brew install jq on macOS).
Copy subagent-statusline.sh to ~/.claude/subagent-statusline.sh and make it executable:
cp subagent-statusline.sh ~/.claude/subagent-statusline.sh
chmod +x ~/.claude/subagent-statusline.shAdd this key to your ~/.claude/settings.json:
"subagentStatusLine": {
"type": "command",
"command": "~/.claude/subagent-statusline.sh"
},Start a new Claude Code session and ask it to spawn a subagent, for example:
Create a sub-agent that echoes "hello world", sleeps for 5 seconds, and then
says "I'm done waiting". This subagent should be run with a haiku model.
While the subagent runs, you'll see its model in the status line at the bottom of your terminal.
Claude Code pipes JSON describing the running subagent tasks to the command configured in subagentStatusLine. The script uses jq to transform each task into a {id, content} object, where content is the text rendered in the status line. That means you can customize the format however you like — reorder fields, add emoji, drop the token count, etc.