AThevon/TokenEater
Native macOS app to monitor AI usage limits and watch your coding sessions live
What it solves
TokenEater provides a real-time monitoring system for users of Claude (Pro, Max, or Team plans) and OpenAI Codex to track their token usage limits. It prevents users from hitting their usage caps blindly by providing visual alerts and pacing data on the macOS desktop.
How it works
The app functions as a native macOS utility that reads OAuth access tokens from the macOS Keychain (for Claude Code) and local auth files (for Codex CLI). It makes read-only API calls to the respective providers to fetch current usage stats and plan information. This data is then presented through several interfaces:
- Menu Bar & Popover: Live percentages and customizable dashboards.
- Desktop Widgets: Native WidgetKit gauges and progress bars.
- Dashboard: A detailed window featuring history sparklines and pacing-vs-equilibrium graphs.
- Agent Watchers: A floating overlay that tracks live Claude Code sessions across terminals and editors.
Who it’s for
Developers and power users on macOS who use Claude Code or OpenAI Codex and need to keep a close eye on their token quotas to manage their workflow without interruption.
Highlights
- Multi-Provider Support: Tracks both Claude and OpenAI Codex side-by-side or independently.
- Smart Pacing & Color: Uses color-coded thresholds and "zones" (chill, on track, warning, hot) to warn users based on burn rate.
- Usage History: Analyzes local session logs to provide stacked charts by model, project rankings, and cache hit rates.
- Native macOS Integration: Fully integrated with WidgetKit and the macOS menu bar.
- Agent Session Tracking: Ability to jump directly to active Claude Code sessions in various terminals (iTerm2, tmux, Kitty, etc.).
Related
- Project
- Project
- Project
- Project
- Project