> demo crew · generated data · sign in to start yours
> percentages compare against the 365 days before this window
this year · ratios of the counted totals
of the context Claude Code sent was read back from the cache instead of sent fresh
> output this year: about 239 novels of text (21.5M words, estimated at 0.75 words per token and 90K words a novel)
last 52 weeks · Monday to Sunday
this year
share of output tokens per model family · last 26 weeks
this year · the reasoning effort each reply ran at
72% of replies
| level | output per reply |
|---|---|
| not set | 679 |
| low | 642 |
| medium | 597 |
| high | 631 |
this year · tokens that ran in subagents (Task, agents) rather than your conversation
2,017 of 39.9K replies in subagents
> counted since Jan 2026: earlier days were sent by a CLI before 0.7.0
this year · replies per block of your computer's own clock · early bird
> counted since Jan 2026: earlier days were sent by a CLI before 0.7.0
this year · cache reads are context Claude Code sent again
output tokens this year by model
Claude Code replies per day · older half on top, newer half below
> counted since Jan 2026: earlier days were sent by a CLI before 0.7.0
average Claude Code replies per weekday this year
> weeks run Monday to Sunday · all time · messages and tool calls count from CLI 0.7.0
