First launch

The setup window connects your providers in about two minutes. Then TokenEater lives in the menu bar.

Setup

  1. 1

    Open TokenEater from Applications. The setup window opens.

  2. 2

    Under Which providers do you use, set up at least one. Claude: click Authorize if offered, then Always Allow when macOS asks (only once). OpenAI: your Codex CLI login is detected with no prompt.

  3. 3

    Under What TokenEater does with them, allow notifications (Send test checks them) and, if you like, turn on Agent Watchers (Claude only).

  4. 4

    Click Finish, active once one provider is ready. Your usage percentage shows in the menu bar: click it for the popover.

Provider not detected

  • Claude: install Claude Code, run claude once to sign in, then click Re-detect.
  • OpenAI: run npm i -g @openai/codex, then codex login, then click Re-detect.

Don't use one? Turn it off with its card's tracking switch (at least one stays on). See Claude and Codex.

After setup

  • Settings -> General -> Replay onboarding runs the setup again. What's new, next to it, replays the 6.0 tour.
  • Codex turns on by itself on first launch (or the first launch after updating to 6.0) if a valid ChatGPT login is found. Signed in later? Use Settings -> Providers, or Track Codex when Monitoring offers it.
  • To start with your Mac: Launch at login in Settings -> General, plus Launch in background to open straight into the menu bar.

Add a widget

Right-click the desktop -> Edit Widgets, search TokenEater, drag a widget out. Codex only? Use Codex Usage or Tokens History. Full list in Widgets. Stale after an update? Remove the widget and add it again.