cc-statusline

A minimal, opinionated and non-configurable, Claude Code status line

Lightweight Claude Code status-line monitor with switchable styles, themes, and slash commands. v3.14.1 lowers the projection red line to 85% (yellow now 70–84) so a 7d window heading for →99% reads red instead of merely warm. v3.14.0 colours the 5h/7d bars by where usage is HEADED, not where it is now: once an end-of-window projection exists, the bar fill, label, and reset clock take their severity from the projected % against the cap (green <80, yellow 80–99, red ≥100) while the fill length still shows current usage — so a 7d window at 24% but on track for →96% reads yellow instead of a falsely-healthy green (all three styles). It also fixes the →NN% projection reading far too low for the first ~15 min after a reset (it was seeded from the used=0 first tick and lagged); it now holds `→--` until MIN_ELAPSED, then seeds from the first trustworthy reading. v3.13.3–v3.13.5 fix the 5h/7d bars sticking at a stale high % after Anthropic re-baselines usage mid-window (e.g. weekly limit raised): the cross-session merge now accepts an official downward revision once the old reading goes unconfirmed for 120s, idle windows replaying hours-old rate_limits blobs (expired 5h reset) can no longer overwrite or re-confirm the shared reading, and the →NN% projection drops its old-denominator samples and relearns after a re-baseline instead of freezing for the rest of the window. v3.13.0 adds a ⚙ session-mode line — effort (low/medium/high/xhigh/max/ultracode/auto) · thinking · fast · output-style — with a distinct static colour gradient per effort tier. v3.12.0 adds an always-on end-of-window projection (→NN%, learned from your local work rhythm) + an at-risk `⚠eta` forecast after each reset timer, session lines on by default, and a faint on-bar version that shows a `↑` hint when a newer release is on PyPI. v3.10.0 adds an opt-in live-activity line (in-progress todo, active tool, completed-tool rollup), git ahead/behind + session duration/lines on the project line, running-subagent lines, an opt-in `bar_shimmer` starfield on the battery bars, and a self-hosted plugin marketplace — plus a reliability fix so auto-update actually runs in daemon mode (detached, non-blocking) and the cache countdown is now per-session-correct. v3.9.0 makes the `cache` countdown auto-detect the real prompt-cache TTL (5m vs 1h) from the transcript, fixing a ~55-minute early `cache COLD` for Claude subscription users (Claude Code requests the 1-hour TTL automatically). v3.7.0 defends against status-line hijacking: a new `cs --setup --project [PATH]` writes a project-level override so another tool stealing the global `~/.claude/settings.json` slot can't kick `cs` out of this project; when displacement is detected, the bar also surfaces a one-line warning. v3.6.0 makes daemon (fast) mode the default for `cs --setup` — under 1% CPU continuously instead of ~3% inline at refreshInterval=1; pass `--inline` to opt back. Requires the `cs` CLI from PyPI (`pip install claude-statusbar` or `uv tool install claude-statusbar`).

Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.

Memory compression system for Claude Code - persist context across sessions

Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.

Curate auto-memory, promote learnings to CLAUDE.md and rules, extract proven patterns into reusable skills.