nanako0129/tokenbar

TokenBar is a macOS menu bar app that monitors how many AI tokens you are using across more than 25 different AI coding tools. Tokens are the unit AI models use to measure text input and output, and each token costs a small amount of money (or consumes a portion of a subscription quota). If you use tools like Claude Code, Codex, Cursor, or Gemini CLI regularly, this app tracks your usage by reading local session log files on your machine, with no data sent anywhere externally. The menu bar icon shows a small summary at a glance: today's token count, estimated cost, tokens-per-minute rate, or how much of your subscription window remains. The remaining quota can be displayed as signal bars, a ring, or a popsicle shape that gradually empties. There is also a running cat animation that moves faster the more tokens you are burning, an idea traced back to the RunCat app. Clicking the icon opens a popover dashboard with six views: a breakdown by AI model ranked by cost, a daily usage calendar, an hourly chart showing when during the day you use tokens, a summary of statistics and usage streaks, a ranking of AI sub-agents by cost, and settings. The dashboard also includes an interactive 3D graph displaying a full year of usage history that you can rotate. Keyboard shortcuts control navigation between tabs and toggle views. The app is built in two layers. A Rust library handles all the data work: parsing session logs from the supported tools, deduplicating entries, calculating costs based on per-model pricing, and fetching subscription quota information. Swift wraps that library and handles everything visual: the menu bar item, the SwiftUI interface, and automatic updates via Sparkle. Building from source requires running make to compile the Rust library first, then building the Swift package. Installation is via Homebrew with a single command. The app requires Apple Silicon and macOS 14 or later, macOS 26 is needed for the full visual style. It is MIT licensed.