AI Token Meter shows the tokens your coding agents use and what they would cost at API prices, right in your Mac menu bar. Today's tokens and their cost sit on top; under them Claude's 5-hour window and Codex's own limits, the sessions working right now, and your history for yesterday, this week, this month or all time, by session, model or project. It reads Claude Code, Codex and Antigravity CLI from the logs they already write, in folders you choose: only token counts, never your prompts or the answers, and nothing leaves your Mac. Prices come from the public LiteLLM table once a day.
Questions
Which coding agents does it read?
Claude Code, Codex and Antigravity CLI. You connect each in Settings by choosing its folder in the macOS open panel, and the meter reads the logs that agent already writes.
Is the cost what I pay?
It is what the tokens would cost on the pay-as-you-go API of each model’s maker. Claude Pro and Max and ChatGPT plans don’t charge per token, so the number is the value of the work done, not your bill.
Does it read my prompts or code?
No. It takes only the token counts, the model, the time and the project folder of each request. It never reads prompts, answers or tool calls, writes nothing into the agents’ folders, and sends nothing anywhere.
Where do the prices come from?
From LiteLLM’s public model price list, the one codigrate.com’s AI Cost Calculator shows. The app downloads it once a day, and you can turn that off.
