Claude Usage Monitor

Starting…

Running now

Waiting for the first reply from a Claude Code session…

All accounts

Account Status 5-hour session Weekly Figures as of Actions

Tracking more accounts

Anthropic reveals an account's usage only to that account's login. Give each extra account its own Claude Code profile — a separate config folder with its own login — and it is tracked alongside the one your Claude Code uses. A switch signs nothing out: the chosen account's login moves into your Claude Code, and the one it replaces moves to a profile of its own, so every account stays signed in here.

  1. Ask Cremind to add a Claude account, e.g. “add a Claude account named spare1”. A window opens with Claude Code in the new profile.
  2. Sign in there with that account, then /exit. Its official figures appear within seconds, and it joins Your Claude Code uses above: one click switches to it.
  3. That's all: the monitor keeps every login renewed while it is idle. Only if an account shows “sign in again”, press Open below and run /login. Switch with the button rather than /login in VS Code: /login drops the account it replaces from this computer.

Activity

Settings: switching and alerts

Either limit interrupts a session when it runs out, so each has the same two alerts. An account past a switch-now threshold isn't suggested as the next one until that limit resets. With automatic switching on, the account your Claude Code uses gets no heads-up alerts: it is switched instead.

Automatic switching
%
%
Used when the switch above is on Automatic. An account whose week is full is set aside until its week resets. Also switches before the thresholds, to an account whose leftover weekly allowance would otherwise go unused within 24 hours.
Accounts it may suggest and switch to
5-hour limit
%
%
Weekly limit
%
%
Higher than the 5-hour ones because 1% of the weekly limit is about 5× as much usage: the last 5% still lasts around half an hour of heavy use.
minutes away
Based on the measured burn rate. 0 turns this off.
Asks Anthropic for each account's usage with its Claude login, the way Claude Code's /usage does — the figures claude.ai shows. Off: estimates between Claude Code's own readings. Each alert is a Cremind skill event. Ask Cremind to notify you about Claude usage alerts and it subscribes to them; with no subscription an alert only appears here.
How the figures are measured

Official figures from Anthropic — the same as claude.ai's Settings → Usage, including usage on claude.ai, the apps and other computers. The monitor asks Anthropic's usage API with each profile's Claude login, as Claude Code's /usage does.

Claude Code's own readings count too: its usage cache, the terminal status line after each reply, and limit-reached notices.

Estimates fill in only when no recent official figure is at hand (live figures off, a profile without a usable login, no network): the monitor adds up what each reply cost at API list prices, from the session transcripts, and converts it with a calibration: … Figures built this way are marked ≈.

Each profile's Claude login is sent only to Anthropic, to ask for its usage (and renew it when it expired, as Claude Code does); it is never copied, logged or shown — a switch moves it between Claude Code's own folders on this computer. Conversation content is never read.