Get started

Everyday cost views

Your real spend, read from the statusline behind your bill.

today · week · month billing-grade

real cost, tokens, and model breakdown for the day / last 7 / last 30 days. Flags: --json, --csv, --clipboard, --since/--until, --group-by <project|branch|cost_center|task|device>, --currency <ISO>.

compare billing-grade

the reveal — your billing-grade cost next to the session-log estimate, and the gap. --days <n> widens the window; --share saves a shareable card to ~/Desktop.

report billing-grade

a monthly cost-center report: wtclaude report --cost-center. --month <YYYY-MM> picks the month.

session [id] billing-grade

list sessions (newest first), or per-turn detail for one id. Supports --json, --group-by, --csv, --clipboard.

project [hash] billing-grade

usage for one project (by hash), or a list of known projects. Supports --since/--until, --json, --csv.

blocks billing-grade

usage in fixed 5-hour UTC buckets on a set grid — not your limit window (for that, see limit). --limit <n> caps how many show.

limit billing-grade

your overall plan limit — % used + reset countdown, from Anthropic’s shared rate-limit bucket (it spans Code, Cowork, and Chat), so it’s your overall plan limit, not a Code-only number.

watch

a live terminal monitor: today’s cost, the current UTC 5h-block burn, and the plan-limit gauge with a basic, non-predictive time-to-limit. --once renders one frame; --interval <s> sets the refresh.

Models & forecasting

What your usage would cost on each model — every projection a labeled estimate.

compare-models estimate

new — re-prices your recorded usage across Opus 5.5, Sonnet 5, and Fable 5.1: what would this month cost on each? Split by surface (Code billing-grade · Cowork a labeled estimate · Chat excluded). It re-prices your recorded usage, not the same task run on each model. Free. --days <n>, --json.

whatif estimate

quick plan- and model-comparison math on your recent usage. Flags: --plan [plan] (on its own, every plan; or one of pro, max_5x, max_20x, team_standard, team_premium), --model <haiku|sonnet|opus|fable>, --days <n>.

fable estimate

reads your billing-grade Fable cost in the terminal and forecasts your pace at Fable list rates. Covers Fable 5.1 (Claude Code’s default Fable model since v2.1.257) and Fable 5. Fable is permanent and plan-conditional: on Max / Team Premium / Enterprise Premium it’s included up to 50% of your weekly usage limit (a share of that limit, not an allowance on top of it, and not a credits wallet); on Pro / Team Standard it bills usage credits from the first token, at $10/$50 — cache-read is $0.25 on Fable 5.1 and $1 on Fable 5, the one price that separates them. A labeled estimate — at standard API list rates; bundle discounts up to 30% and promos not reflected. --days <n>, --plan <plan>, --json.

forecast estimate

projected Agent-SDK spend from your recorded usage; if a revised split lands, it compares that against the credits the split includes. (The June-15 credit split it was built for was paused.) --days <n>, --json.

credits estimate

your fast-mode credit spend this month (billing-grade when Claude Code reports it, otherwise labeled inferred); the Agent-SDK credit pool shows as paused. WTClaude can’t read your usage-credit balance yet — check it in Claude’s usage settings. --json.

readiness estimate

a one-time credit-coverage check (built for the June-14 readiness window; the split was paused, so it reads as ready). --json.

Opus 5.5 is Claude Code’s default model since v2.1.280 (claude-opus-5-5). WTClaude prices it correctly — $4/$20, with cache reads at $0.20 per million, off the model’s own reported tokens, with no unknown-model gaps. Sonnet 5 stays at $2/$10 — the increase to $3/$15 once scheduled for September 1, 2026 was cancelled.

Insights

Beyond the dollars — where your context and effort actually go.

waste estimate

new — reveals the always-loaded skills and subagents you never invoke, and what re-reading them costs, computed at list cache-read rates (each item’s token size is a labeled estimate). CLAUDE.md files are listed as always loaded but not judged. A read-only reveal: it recommends a REVIEW, never removes or changes anything. --days <n>, --all, --json.

quality

your one-shot success rate — edits the model landed without a retry (file-aware, salted). The command ships; the edit-target field it reads isn’t in current Claude Code payloads yet, so today it reports “not available yet”. Supports --since/--until, --json, --csv.

debrief

an end-of-day summary — the day’s total with its cost basis, your costliest turn, and the cache-read share of your recorded input-side tokens (context occupancy, not a hit rate).

tasks coming soon

token/cost breakdown by task category — coming soon (the live payload doesn’t expose tool names yet, so categories aren’t populated).

badges

the usage badges you’ve earned.

leaderboard

your local leaderboard standing, plus a privacy preview of exactly what a share would expose. --json.

Sync & dashboard

Opt-in and local-first — nothing leaves your machine until you turn it on.

sync

opt-in cloud sync + multi-device totals. --status / --enable / --disable. Local-first — nothing leaves your machine until you enable it, and you see a preview of what would sync before it does.

dashboard

open the web dashboard in your browser — everything the CLI tracks, in one view.

devices

per-device usage from this machine’s session files — always a local-only view; with sync on, cross-device totals are on the dashboard. Supports --since/--until, --json, --csv.

share

manage anonymous data sharing for the leaderboard + benchmarks. --preview shows exactly what would be shared; --enable / --disable.

invite

generate a shareable invite link.

Setup & utility

setup

configure WTClaude — wires the Claude Code statusline and creates your local data dir — no login, no account. --plan <plan> (pro, max5, max20, team_standard, team_premium, enterprise_standard or enterprise_premium), --yes for non-interactive.

export

dump all local WTClaude data to JSON — everything stored about you. --out <file>, --pretty.

uninstall

remove the WTClaude statusline hook. --keep-data leaves your local data; --purge also deletes ~/.wtclaude/.

statusline

render the WTClaude status line from your latest turn — a preview of what the collector prints back into Claude Code.