Skip to content

v1.1.2

Latest

Choose a tag to compare

@Tr1pl3x Tr1pl3x released this 25 Jul 07:04

A live terminal dashboard for your Claude plan limits — run it once when you
clock in and never glance at the in-app /usage screen again.

Fixed in v1.1.2

Update if you're on any earlier 1.1.x. Two more causes of the "rate
limited — backing off"
screen, both about the dashboard asking too often:

  • Only one dashboard can run at a time now. Anthropic's rate limit is per
    account, not per window, and the usage endpoint only has room for about one
    request every couple of minutes. Two copies polling the same login will
    rate-limit each other however politely each behaves alone — and a second copy
    is easy to end up with by accident. It now says "already running" and closes
    instead of competing.
  • Being rate-limited no longer makes it poll faster. The first retry after a
    429 was a fixed 60 seconds while the normal poll is 120 — so hitting the limit
    doubled the request rate until the back-off grew. Retry waits are now floored
    at whatever your poll interval is.

(v1.1.1 fixed the same class of problem in the R key: its cooldown was 5s,
and a refused press gave no feedback, so people pressed it again. It's now 30s
with a visible [R] in 22s countdown.)

What's new in v1.1.x

  • Per-model usage tracking. Weekly per-model caps now get their own bars
    (Fable, Opus, …) under the overall 7-day window — the same breakdown the
    in-app usage screen shows. These had silently stopped appearing: Anthropic
    moved the data into a new field, and the dashboard was still reading the old
    one.
  • A "nearest limit" line naming whichever window is furthest along. Worth
    watching — a per-model cap is often further along than your overall weekly
    bar, so it's the one you'll actually hit first.
  • Press R to refresh on demand.

(Earlier 1.0.x releases added live-ticking reset countdowns, a tunable poll
interval, automatic back-off for rate limits, and a clean single-screen error
state.)

Download

Grab claude-usage.exe below and double-click it. No Python or install needed.

Requirements

  • Logged into Claude Code with a Claude Pro or Max subscription on this
    machine (it reads the OAuth token Claude Code already stores).
  • Windows. (macOS keeps the token in Keychain — not supported yet.)

What it shows

  • 5-hour, weekly (7-day) and per-model usage with live, colour-coded bars
  • Live reset countdowns and extra-usage credits
  • Fetches every 120s (or press R); strictly read-only — never touches
    your login

See the README for details and known limitations.