Token use, run by run

Mean tokens per chore in each daily run, per model, as each CLI reports them, on today's chores only, then each chore on its own. Hover a chart (or focus it and use the arrow keys) for the numbers. The change is against the run before, over the chores both runs recorded. A run with no usage record is not counted as 0: it stays out, and the page says how many.

API $ is what the same tokens would cost on each provider's pay-as-you-go API at its list price (checked 2026-11-14): Anthropic, OpenAI, Google. The runs themselves use subscription plans, so this is a scale, not a bill: cache reads at the cache price, reasoning at the output price, Claude cache writes at the 5-minute price, OpenAI at the short-context price, and every run at the price of its own day, so a price change never reprices past runs. Chores with no usage record or no known price are left out, so totals read low there. Per agent and model only.