Blog · · 5 min read

How to check where you stand on Claude's weekly limit

Anthropic ships the data. Finding it takes more clicks than it should. Here's where the page actually lives, what each number means, and a small free tool that puts the answer in your menu bar — alongside Codex and Gemini.

If you've been on Claude.ai's paid plan long enough to hit the weekly wall, you already know the feeling. You're mid-prompt. Claude tells you you're out of weekly capacity. You stare at the message wondering when, exactly, your week resets — and you go looking for the answer.

The page exists. Anthropic does publish the numbers. They're just buried more deeply than they should be. This is a short guide to finding them, understanding them, and (optionally) skipping the click-through entirely with a free tool that surfaces them in your macOS menu bar.

Where Claude's usage page actually lives

The fastest path:

  1. Go to claude.ai and sign in.
  2. Click your avatar (bottom-left on desktop, top-right on mobile).
  3. Pick Settings.
  4. In the Settings sidebar, choose Usage.

You'll land on a page showing two numbers: a 5-hour usage figure and a weekly usage figure, each as a percentage of your plan's allowance, each with a reset timestamp. If you're on Claude Pro, the weekly cap is shared across the conversational Claude.ai web app and Claude Code; if you're on Claude Max, the caps scale up but the same shape applies.

What the two numbers actually mean

Anthropic uses two overlapping windows. They look similar at a glance and they aren't.

The 5-hour rolling window

The 5-hour window resets continuously — five hours after your first message, on a sliding basis. It's the limit you're most likely to hit in a single coding session. If you're paired with Claude Code on a substantive feature, you can fill this faster than you'd guess, especially with long context windows in play. The reset time shown on the usage page is the moment your current 5-hour bucket will roll over, not a fixed clock.

The weekly allowance

The weekly cap resets at a fixed time once every 7 days. Anthropic picks that time when you start your subscription and it stays stable from there. The weekly is the harder one to track informally because it accrues silently across a whole week of work; by the time you notice you're 90% through, you're often two days from reset and rationing.

Why the page is opaque on first read

The two numbers are sitting next to each other but they aren't directly comparable. A 70% 5-hour usage and a 70% weekly usage have very different implications for the rest of your day. One refills in hours; the other in days. Reading them as "I'm at 70%" without separating which window the figure refers to is the source of most of the wall-hitting that happens on Wednesday afternoons.

The cross-provider problem

If you only use Claude, the page above is enough. Click through, glance at the numbers, close the tab. Most developers don't use only Claude.

A typical AI-tools-heavy workflow in mid-2026 looks more like this: Claude Code for one project, Codex CLI for another, Gemini for a third. Each has its own subscription, its own limits, and — critically — its own way of showing usage. Codex doesn't ship a built-in usage UI worth opening; checking it means digging through ~/.codex files or visiting the OpenAI billing dashboard. Gemini's usage page is a separate URL on Google's domain with its own sign-in.

Knowing where you stand across all three becomes a three-tab, three-sign-in, three-mental-context exercise. By the time you've finished checking, you've lost the thread of whatever you were doing.

The shortcut: a free menu bar app

This is the small problem that became Model Meter — a free macOS menu bar app that consolidates Codex, Claude, and Gemini usage into a single readout, sitting next to the clock.

The app shows you a line like C 8% · Cl 5% · G 1% at all times. Click it for the full per-provider breakdown — both the 5-hour rolling window and the weekly allowance, side by side, with their reset times. No tabs to open, no sign-ins to navigate.

A few details that matter:

  • Free and open source under MIT. The source is on GitHub.
  • No telemetry. The app phones home for one thing: checking the update feed. Nothing about your usage or accounts is sent anywhere.
  • No API keys. Model Meter reads your existing authenticated sessions (Claude.ai's session via Keychain, Gemini's via an embedded WebKit view, Codex's via the local rate-limit files Codex already writes). Your OpenAI / Anthropic / Google API spend is unaffected.
  • macOS 14 or later. Signed and notarized by Apple. Sparkle handles auto-updates from here on.

Get Model Meter

If you'd rather just bookmark the pages

For the resistant-to-installing-things, here are the canonical usage URLs across the three providers:

Bookmark the three. Keep them in a folder. Check them when you're about to start a long session, and again when something feels suspiciously throttled. It's perfectly workable; it's just three tabs and three sign-ins more than necessary.

Either way — the answer to "where am I against my Claude weekly limit?" is reachable. Now you know where the page is and what the two numbers mean. The rest is just how many clicks you want to spend on the question.