TokenBar

Token usage, estimated cost and rate limits for your AI coding tools — in the macOS menu bar.
Reads local logs only. No API keys, no accounts, no telemetry.

Features

• Usage at a glance — tokens and list-price cost for Today / 7d / 30d / All, split into input, output, cache read and cache write, with a 14-day sparkline per tool.
• Rate limits — Claude 5-hour and weekly windows (incl. per-model weekly), Codex 5-hour and weekly, with reset countdown and clock time.
• Where it came from — usage attributed to source (Terminal, Claude Desktop, Zed, VS Code, …) and to project folder.
• Live sessions — conversations active in the last 15 minutes with project, source, model, calls, cost, duration and a context-window gauge (how full the model's context is); notification at 80 % so you can /compact before it auto-compacts.
• Dashboard window — 30-day stacked chart, top projects / models / sources; click a bar to drill into one day.
• Alerts — limit thresholds (80 % / 95 %), daily and weekly spend budgets, and a break reminder after long continuous use.
• Menu bar, your way — combined total or a single tool (its logo + tokens + worst limit); amber / red when a limit is close.
• **Exp