Read your plan usage

Agents: claude, codex, muse · On: phone, desktop

A row of meters above the transcript shows how much of each logged-in plan's quota you have spent, and when each window resets. Providers that are logged out have no meter card.

Why you would use it

Before starting a long run it is worth knowing whether you have four hours of quota or four minutes, and which agent still has room.

How to use it

1. Look at the meter row (#agents) above the plan strip. One card per logged-in provider.
2. The card of the agent this tab runs is highlighted.
3. Tap the latency segment in the status bar to force a refresh.

When every provider is logged out, the whole row collapses:

The typed /usage on claude is answered by VCode rather than forwarded: the meters above the transcript show your plan usage.

What you see

Each card carries the agent's mark and one row per reported window:

• the window's name, derived from its length: 5h, W for the weekly window, or 2d/45m for anything else
• the percentage used
• a bar, coloured by how full it is, with the weekly bar drawn slightly dimmer
• how long until it resets: 42m, 3h 20m, 2d 4h, or now

With no windows reported the row reads · and two dashes. The card's tooltip is the agent's name and the plan, for example Claude · Max.

Limits and known gaps

• No row for opencode: it reports no account usage windows.
• No row for the test-only Mock agent; see the-three-agents.
• Figures are cached for ten minutes by default, so a meter can lag what the CLI just spent.
• If a refresh finds no provider available, the previous figures are kept and marked stale rather than blanked.
• When a provider cannot be read at all, its label falls back to GPT -- or Claude -- with the reason attached.
• Per-session dollar cost is not shown. /cost on claude is answered with per-session spend is not surfaced in VCode yet.

Related

• usage-limit-pause-and-retry: what happens when a window is full
• status-bar: the per-tab context figure, which is a different number