Configure Claude Code status line to show token usage
Claude Code status line — token counter setup
Motivation
No counter = no signal for dumb zone. Past ~75K tokens (HumanLayer) / ~100K rounded (Pocock) = quality drop. See [[context-cliff-100k]]. Status line = ambient visibility, zero prompt cost.
Setup
1. Edit ~/.claude/settings.json
"statusLine": {
"type": "command",
"command": "/c/Users/C5420321/.claude/status-line.sh"
}
2. Write script ~/.claude/status-line.sh
POSIX, Git Bash compatible on Windows:
#!/usr/bin/env sh
input=$(cat)
model=$(echo "$input" | jq -r '.model.display_name // "?"')
tokens=$(echo "$input" | jq -r '.context.tokens // 0')
dir=$(basename "$(echo "$input" | jq -r '.workspace.current_dir')")
branch=$(git -C "$(echo "$input" | jq -r '.workspace.current_dir')" branch --show-current 2>/dev/null || echo "-")
warn=""
[ "$tokens" -gt 75000 ] && warn="⚠"
printf "%s · %s · %s · %s/75K%s" "$dir" "$branch" "$model" "$tokens" "$warn"
chmod +x ~/.claude/status-line.sh.
3. Test
Run /statusline in CC session. Renders inline. Format:
| Field | Source |
|---|---|
| dir | workspace.current_dir basename |
| branch | git branch --show-current |
| model | model.display_name from stdin JSON |
| tokens | context.tokens from stdin JSON |
| warn | ⚠ past 75K |
Alternate: built-in skill
Skip manual script. Invoke statusline-setup skill — auto-writes config + script.
Refs
- Matt Pocock, "Creating The Perfect Claude Code Status Line" — https://aihero.dev
- Anthropic docs — https://docs.anthropic.com/en/docs/claude-code/settings#status-line
- Source: Matt Pocock workshop 2026-07-03
Cross-refs
- [[context-cliff-100k]] — why the counter matters
claude-code-latency-techniques— companion CC settings- Built-in Claude Code skill:
statusline-setup