尚未收集仓库信息。
尚未检测到更新。GitHub 发布版本后会显示在这里。
A feature-rich status line for Claude Code with colors, rate limits, git info, session cost, worktree detection, and auto-compact warnings
翻译准备中
正在加载完整 README。
查看完整 README ↗A feature-rich status line for Claude Code that displays model info, context usage, git status, rate limits, session cost, and more — quiet when everything is healthy, loud exactly where something needs attention. Zero-config by default, deeply customizable when you want it (thresholds, palette, glyph sets, per-segment toggles).
📖 Full documentation — installation · usage · configuration · tips · troubleshooting
Click to play, or watch it on the documentation site.
| Segment | Description |
|---|---|
| Model | Short model name (e.g., Opus 4.6 instead of Claude Opus 4.6) |
| Context | Usage percentage + token counts (e.g., 12% (42k/200k)) with color coding |
| Auto-compact warning | Steady bold ⚠ when context usage >= 80% |
| Directory | Truncated to last 2 path components |
| Git branch | Branch name with detailed status: staged (S:2), unstaged (U:1), untracked (A:3) |
| Git ahead/behind | ⇡2⇣1 arrows showing commits ahead/behind upstream |
| Worktree | ⎇wt indicator when running inside a git worktree |
| Session cost | $X.XX with configurable warn/critical thresholds and currency |
| Session duration | 5m, 1h30m, etc. |
| Lines changed | +156/-23 cumulative lines added/removed |
| Per-turn cost | 48k/turn — what the next message costs before you type (at the current model's cache-read rate), and 88t turns so far |
| Cold prompt cache | 486k next turn (cache cold) — once the prompt cache has expired, the next message re-writes the whole context |
| Effort level | ● high, ◑ medium, ◔ low (from the session itself) |
| Badges | Subagent name, fast mode, thinking, vim, non-default output style, ⚠200k+ |
| Rate limits | █░ progress bars for 5-hour, 7-day, per-model, and extra usage with reset times |
| Pace projection | ⇢ 94% — where the 5-hour / 7-day window lands at the current burn rate |
| Context hint | One line naming the right command: /compact for a full window, /clear for an expensive one |
One escalation scale drives every percentage-based segment (context, rate bars) — quiet when healthy, then:
CESL_WARN)CESL_HIGH); context also gains a steady bold ⚠CESL_CRIT)Session cost uses its own CESL_COST_WARN/CESL_COST_CRIT thresholds. The per-turn cost uses absolute token thresholds (CESL_CTX_WARN/CESL_CTX_HIGH) because a percentage of a 1M window says nothing about what a message costs — 486k reads as a comfortable 49% full while re-sending 48k tokens every turn at a 10% cache-read rate (24k on Opus 5.5). The model name is colored by model family (Opus, Sonnet, Haiku, Fable).
NO_COLOR and TERM=dumb are honored; CESL_COLOR=0/1 overrides the detection either way.
jq — JSON parsingcurl — fetching rate limit data from the Anthropic APIgit — branch and worktree detectiongit clone https://github.com/dsebastien/claude-epic-status-line.git
cd claude-epic-status-line
bash install.sh
statusline.sh to ~/.claude/statusline-command.sh~/.claude/settings.json:{
"statusLine": {
"type": "command",
"command": "bash '/home/YOUR_USER/.claude/statusline-command.sh'"
}
}
bash uninstall.sh
Or manually remove the statusLine key from ~/.claude/settings.json and delete ~/.claude/statusline-command.sh.
Everything is optional — with zero config the status line renders its default design. Settings layer as: script defaults < ~/.config/claude-epic-status-line/config.sh < CESL_* environment variables (environment wins).
以上为 README 的开头部分。其余内容请阅读原文。 查看完整 README ↗