Estimated weekly usage racing the clock. When ⚡ Usage trails ⏱ Time, that striped stretch is headroom you've paid for.
Background/loop sessions add up fast. Keep long runs intentional; kill idle ones.
Each subagent runs its own requests. Spawn deliberately; use a cheaper model for simple ones.
Longer context costs more even when cached. /compact mid-task, /clear when switching tasks.
All sessions share one weekly limit. Queue work you don't need all at once.
| Fed | Session $ | Week % |
|---|---|---|
| 7:28am CT · Jul 11 | $147.09 | 27% |
| 12:07am CT · Jul 11 | — | 26% |
| ~11:21pm Jul 10 | — | 26% |
| 7:38pm CT · Jul 10 | $115.04 | 25% |
| 7:28pm CT · Jul 10 | $113.74 | 24% |
| 7:02pm CT · Jul 10 | $109.48 | 24% |
| 6:07pm CT · Jul 10 | $102.43 | 22% |
| 5:48pm CT · Jul 10 | $95.55 | 21% |
| 5:42pm CT · Jul 10 | $94.43 | 20% |
| 5:21pm CT · Jul 10 | $92.54 | 20% |
| 5:19pm CT · Jul 10 | $90.69 | 20% |
| 4:30pm CT · Jul 10 | — | 19% |
| 3:18pm CT · Jul 10 | $83.58 | 18% |
| 2:47pm CT · Jul 10 | $69.93 | 13% |
| 2:43pm CT · Jul 10 | — | 11% |
| 2:27pm CT · Jul 10 | $63.11 | 9% |
| 2:03pm · Jul 10 | $4.41 | 8% |
| 10:45am CT · Jul 11 | $157.10 | 28% |
The big number is an estimate: ccusage measures this machine's API-equivalent spend, calibrated to your last real reading (27% at 7:28am CT · Jul 11). It updates hands-free from ccusage — no pasting. It counts this machine only, so paste /usage occasionally to re-calibrate against the true cross-device total. ccusage last read: ~12:00pm CT Jul 11.