The Usage pane

Open with what’s the usage (or open usage). It answers “how much have my agents been burning?” from local data: each detected agent shows an estimated cost and token totals. The window opens half again as large as it used to, so the figure and its sparkline are readable without resizing it every time. The compact widget on the starter canvas keeps its own smaller size.

Reading it

• Compact view - the headline figures; click to expand.
• Expanded view - one row per agent: cost estimate, token counts, and a note when an agent is installed but has no recorded usage (or isn’t installed at all).

Count jobs on this computer

An opt-in checkbox at the bottom of the expanded widget. When it is on, Mastery HQ counts what happens to every project started from a brief: jobs started, finished and abandoned (no end after a day), the median time from the brief to the first file on disk, which pre-flight issue stopped a run, which error card showed, which kinds of deliverable you opened, and media lines an agent wrote that were not understood. Counts and codes only: no brief text, file name or output is kept, and nothing leaves this computer. Turning it off clears the counts.

Good to know

• Figures are local and estimated - the pane says so up front. Different CLIs log with different fidelity; treat totals as directional.
• Agent windows also show a live per-window token meter in their header - that’s per-conversation, while this pane is per-agent overall.
• For what happened when, the Audit Log is the companion view.