Claude Status

A custom status line for Claude Code that shows everything you need at a glance.

⬡ Claude Opus 4.6   ⌂ ~/…/2026/ClaudeStatus  ⎇ main*
ctx ▰▰▰▰▱▱▱▱ 48% 192k/400k · 5h ▰▱▱▱▱▱▱▱ 8% ↻ 4h 12m · 7d ▰▰▰▰▰▰▰▱ 87% ↻ 1d 8h · ~$4.20

What it shows

Element	Description
Model	Current Claude model (Opus, Sonnet, Haiku)
Directory	~/…/parent/folder — tilde-relative with smart truncation
Git branch	Current branch + * dirty indicator for uncommitted changes
Context	Progress bar + percentage + token count (e.g. 192k/400k)
5h usage	Rate-limit utilization with reset countdown
7d usage	Weekly rate-limit utilization with reset countdown
Cost	Cumulative API-equivalent session cost, tracked per-turn

Visual indicators

• Progress bars change color: green (< 60%) -> amber (60-84%) -> red (85%+)
• Stale cache data (> 5 min) marked with amber ?
• Git dirty state shown as * next to branch name
• Directory prefix (~/…/) shown in muted color, current folder in bold

Requirements

• Claude Code CLI
• jq — JSON processor
• curl — for API usage fetching
• git — for branch/dirty detection

Install

git clone https://github.com/sam-pop/ClaudeStatus.git
cd ClaudeStatus
bash install.sh

Then restart Claude Code.

What the installer does

1. Copies statusline.sh and fetch-usage.sh to ~/.claude/
2. Sets statusLine in ~/.claude/settings.json
3. Adds hooks to fetch usage data in the background
4. Backs up your existing settings to settings.json.bak
5. Primes the usage cache

Uninstall

bash uninstall.sh

Removes scripts, cleans settings, and deletes the cache. Your other Claude Code settings are preserved.

How it works

Status line (statusline.sh): Receives JSON from Claude Code via stdin on every prompt. Extracts model, directory, context window data in a single jq pass. Renders two lines with ANSI color codes.

Usage fetcher (fetch-usage.sh): Runs in the background via Claude Code hooks (on every tool use and on stop). Calls the Anthropic API to get 5-hour and 7-day rate-limit utilization. Results are cached to /tmp/.claude_usage_cache with a 60-second throttle to avoid excessive API calls.

Cost tracker: Detects new turns by watching output token count. Each turn's cost is computed from the full input breakdown (cached reads, cache writes, uncached input) plus new output tokens, then accumulated in /tmp/.claude_cost_session. Auto-resets on new sessions.

Architecture

Claude Code
    |
    ├── stdin JSON ──> statusline.sh ──> rendered status bar
    |                       |
    |                       └── /tmp/.claude_cost_session (per-turn accumulator)
    |
    └── hooks ──> fetch-usage.sh ──> /tmp/.claude_usage_cache
                       |
                       └── Anthropic API (OAuth usage endpoint)

Configuration

Edit ~/.claude/statusline-command.sh directly to customize:

Variable	Default	Description
BAR_W	8	Progress bar width (characters)
CACHE_MAX_AGE	300	Seconds before cache is marked stale
C_MODEL	warm orange	Model name color (RGB)
C_DIR	teal	Directory color (RGB)
C_BRANCH	violet	Git branch color (RGB)
C_COST	sage green	Cost estimate color (RGB)

Cost estimation

Session cost is tracked cumulatively per-turn using Anthropic API pricing:

Model	Input	Output	Cache Read	Cache Write
Opus	$15/MTok	$75/MTok	$1.50/MTok	$18.75/MTok
Sonnet	$3/MTok	$15/MTok	$0.30/MTok	$3.75/MTok
Haiku	$0.80/MTok	$4/MTok	$0.08/MTok	$1/MTok

This shows what your session would cost at API rates. Actual cost depends on your subscription plan.

Platform support

Platform	Status line	Usage fetching
macOS	Full support	Full support (Keychain)
Linux	Full support	Partial (needs credential file)

On Linux, the usage fetcher looks for credentials in ~/.config/claude-code/credentials.json or ~/.claude/credentials.json.

License

MIT