Subscription truth
Loading…
Waiting for the Chrome extension to finish scraping
claude.ai/settings/usage… the page should appear in a
moment.
Mind — new scrape
Plan
—
Observed
—
—
Current 5-hour window
Session used
—%
resets in — min
Mind sees
—%
— tokens this window
Off-device
—%
claude.ai & mobile chats
Weekly
All models
—%
Sonnet only
—%
Claude Design
—%
Calibration — 5-hour window
Effective cap
—
Implied cap (this scrape)
—
Calibration — weekly (7-day rolling)
claude.ai weekly
—%
Mind sees (last 7d)
—
Configured weekly cap
—
Implied weekly cap
—
Off-device this week
—
scraped % − Mind-visible %
Current window — tokens over time
Solid line: Mind-tracked cumulative tokens · Dashed: linear projection to window end ·
Dots: scrape datapoints (claude.ai-reported, converted to tokens via cap).
Projection accuracy — Mind vs scraped
Trend: loading…
Per-scrape variance over time. Bars are signed Δ (scraped% − mind%);
solid line is |Δ|; dashed line is the OLS trend in |Δ| — its slope
drives the verdict above. Flat near zero ⇒ Mind has full visibility.
The detail chart above plots all three signals (scraped, Mind, off-device)
separately at each scrape.
Usage breakdown over time
Stacked bars: Claude Code ·
Mind direct ·
off-device (claude.ai / mobile).
Yellow burn-rate line shows average tokens/hour over each bucket.
Recent observations
| When | Source | Plan | Session | Weekly | Off-device |
|---|
Mind — instance info