CodexBar

CodexBar is a macOS menu-bar app and CLI that surfaces AI coding-provider usage windows, credit balances, reset countdowns, and local token-cost estimates across Codex, Claude, Cursor, Gemini, and many other providers. We adopt it for day-to-day token budgeting and model-switch decisions. Pair with agent workflows where subscription limits and per-model cost vary.

Blurb

Tiny macOS 14+ menu bar app that keeps AI coding-provider limits visible and shows when each window resets. Plan around resets, credits, spend, and cost scans. Privacy-first: reuses existing provider sessions so no passwords are stored.

Summary

CodexBar tracks per-provider session, weekly, and monthly usage windows with countdowns to the next reset. It also runs local cost scans from conversation logs (codexbar cost --provider codex, claude, or both) when API billing detail is unavailable. Estimates are API-equivalent, not exact subscription charges.

Use when you run multiple coding agents and need visibility before starting a long task. The bundled CLI supports scripts and CI on macOS and Linux. Contrast RTK, which compresses shell output before it hits the context window; CodexBar measures spend and limits after the fact.

Details

• Install: brew install --cask steipete/tap/codexbar or GitHub Releases; Sparkle auto-updates on macOS.
• CLI: codexbar cost analyzes local Claude and Codex history offline; returns session and 30-day token and USD estimates.
• Providers: Codex, OpenAI, Claude, Cursor, Gemini, Copilot, OpenRouter, LiteLLM proxy, Vertex AI, AWS Bedrock, and others per upstream docs.
• Caveats: Local cost figures are estimates from published pricing and local logs; flat-rate plans may not match billed amounts.
• Fit: Tool / AI Agent ops layer for cost discipline and model routing decisions.