tokenmeter

Changelog

What changed in each release. Upgrade with brew upgrade tokenmeter or pipx upgrade tokenmeter-dashboard, then tokenmeter start.

  1. 0.3.0
    • Tips to spend less: one click sends your 5 or 10 most expensive prompts to Claude Code, Codex or Copilot CLI on your machine and shows what to change. Or copy the prompt into any AI.
    • Run tokenmeter start once after upgrading so Get tips finds your AI tool.
  2. 0.2.10
    • Clearer setup: tokenmeter start explains when the command is not on your PATH yet, and how to fix it.
    • The Spend tile links to plan setup when no subscription is configured.
  3. 0.2.6
    • Starts at login: Tokenmeter runs its own login service on macOS and Linux, so it starts again after a reboot without Homebrew services.
  4. 0.2.5
    • Smoother upgrades: tokenmeter start restarts an older running copy after an upgrade.
  5. 0.2.4
    • Start and stop: new tokenmeter start and tokenmeter stop. If port 7788 is taken, Tokenmeter uses the next free port, and --port saves your choice.
  6. 0.2.3
    • Instant start: the dashboard opens instantly and indexes large histories in the background.
    • Prices for GPT-6 Sol and Luna, with OpenAI long-context rates.
  7. 0.2.1
    • Cost you can check: tokens are split into cached, fresh and output, with the cost math shown for every prompt.
    • Your own settings live in ~/.tokenmeter/config.json.
  8. 0.2.0
    • First release: on PyPI and Homebrew, with per-prompt cost, usage limits, cache misses, cost per commit, Copilot CLI support, team mode and a menu bar widget.