You find out you have hit your Claude limit in the worst possible way: mid-session, mid-thought, with the work unfinished. Claude Meter fixes exactly that, by putting your five-hour and weekly usage numbers right in your Mac’s menu bar, where you can see them at a glance without breaking your flow.
If you use Claude Code seriously, you already know the shape of the problem. The five-hour session window ticks down whether you are actively coding or just have a session open. The weekly cap compounds this. You might have plenty of five-hour headroom left but still hit a wall because your weekly allowance ran out on an Opus-heavy afternoon three days ago. And because Claude’s rate limits are not displayed anywhere obvious while you work, most people discover them the hard way.
Claude Meter, built by developer Jewei Mak, is a free, open-source macOS app that solves this without ceremony. It is a small, native Swift menu bar pill that shows you exactly where you stand, live, updated automatically, with reset countdowns running underneath each number. No Dock icon. No setup beyond dragging it to your Applications folder.
Why This Even Needs to Exist
Claude’s subscription plans have two separate limit structures running in parallel. Understanding both is the whole reason a tool like this is useful.
The 5-hour rolling window is a rate limit on how much you can use in any 300-minute span. It resets on a rolling basis, not at midnight, not at the top of the hour. Your first message starts the clock, and everything you send in the next five hours draws from that window.
The weekly cap is the harder ceiling. It covers all models across the seven-day period and resets once a week. On Max plans, there is a separate per-model cap on top of this. Spend heavily on Opus or Sonnet in the first half of the week and you may exhaust your model-specific budget while your overall weekly percentage still looks fine.
That last part is the one that actually catches people off guard. The percentage in the menu bar alone does not tell you whether a per-model cap has run out, which means you could be at 65% overall and still find that the model you need is blocked. Claude Meter surfaces all three numbers: the five-hour session percentage, the weekly all-models percentage, and the per-model cap, so you are looking at the thing that will actually stop your work, not just a reassuring headline number.
This is also the problem that motivated the May 2026 doubling of Claude Code’s 5-hour limits. Anthropic doubled session caps across every paid plan after sustained user feedback about hitting the wall unexpectedly. The weekly cap structure stayed the same. Which is to say: the problem got better, not gone.
What Claude Meter Shows You
The menu bar pill is deliberately minimal. It shows the nearest limit as a percentage. At a glance, you know whether you have breathing room or whether you are in the red zone.
Click the pill and you get the full popover:
- Session (5-hour): your current rolling window usage as a percentage, with a live countdown to reset
- Weekly (all models): your seven-day cumulative usage, again as a percentage, with the reset date
- Per-model cap: the model-specific weekly budget that can cut you off before the overall weekly number hits 100%
Color moves from calm to amber to red as you approach each threshold. Optional notifications fire at warning and critical levels so you get a heads-up before the interruption, not at the moment of interruption.
There is also a widget you can drop on your desktop or in Notification Center if you prefer a persistent, always-on view without even needing to click the menu bar.
How to Install Claude Meter (Step-by-Step)
The whole process takes about two minutes. You need a Mac running macOS 14 (Sonoma) or later, and Claude Code signed in at least once on that machine.
- Go to the GitHub releases page. Head to github.com/jewei/claude-meter/releases/latest and download the most recent
ClaudeMeter-[version].dmgfile. - Open the DMG and drag Claude Meter to Applications. Double-click the .dmg file. A window opens showing the Claude Meter icon and your Applications folder. Drag Claude Meter into Applications, exactly the same way you install any Mac app.
- Launch the app. Go to your Applications folder and double-click Claude Meter. No Dock icon appears, which is intentional. The meter lives in your menu bar, top right on your screen. Look for the small percentage indicator.
- Let it detect Claude Code automatically. If you have Claude Code installed and signed in, Claude Meter installs a transparent bridge into your Claude Code statusline configuration. Nothing about your statusline changes; the bridge just passes usage data to the meter. Your real usage numbers appear immediately. No API key required, no manual token setup.
- Click the menu bar pill to see the full breakdown. Click the percentage in your menu bar to open the popover. You will see your five-hour session usage, your weekly usage, and the per-model cap if your plan has one, each with a reset countdown.
- Set your notification thresholds (optional). Click the settings gear inside the popover. You can set custom warning and critical percentages, enable or disable notifications per limit type, and configure auto-launch at login. The defaults are sensible, so if you do not want to tweak anything, you do not have to.
- Enable launch at login. If you want Claude Meter running every time you start your Mac, turn on the launch-at-login toggle in settings. Set it once and forget it.
The build is Developer-ID signed and notarized by Apple, so macOS will not flag it as unverified software. No right-click-to-open workarounds needed.
How It Gets Your Data (And Why This Matters)
The numbers Claude Meter shows come from the same usage endpoint that Claude Code’s own /usage command calls. Specifically, it reads from the Anthropic OAuth usage API, using the credentials Claude Code already stores in your macOS Keychain. Nothing additional is ever sent anywhere.
Here’s the thing that makes this cleaner than most alternatives: Claude Meter does not scrape your browser, parse transcript logs, or estimate usage from local files. The numbers reflect what Anthropic’s systems actually see for your account, refreshed every time the app polls.
Each reading carries a timestamp so you can tell if you are looking at a fresh number or a slightly stale one. Leave your Mac overnight and come back: Claude Meter correctly shows 0% on the session meter once the five-hour window has expired, rather than clinging to yesterday’s number.
For users running multiple Claude Code sessions across different projects, the app reads all of them and merges intelligently, picking the freshest reading per window. The result is one accurate picture, not a guess based on whichever session updated last.
Everything stays local. Credentials stay in Keychain. The diagnostics tool, if you ever need it for a bug report, sanitizes email addresses, home directory paths, and session identifiers before anything is copied out.
Multi-Account and Multi-Provider Support
If you run multiple Claude Code accounts simultaneously, Claude Meter handles this natively. You can set a display name and plan badge for each account, and the popover shows activity rings per account side by side.
Beyond Claude, Claude Meter optionally tracks Cursor billing-period usage, Codex homes, and Grok CLI credits. These non-Claude providers are kept visually separate so they do not clutter your Claude meter or trigger Claude-specific notifications. The menu bar pill stays focused on your Claude numbers.
This matters if you are using Claude Code as your primary driver but occasionally switching to Cursor or Codex for specific tasks. You get a consolidated view of where you stand across tools without having to juggle separate monitoring apps.
Honest Limitations
A few things worth knowing before you install:
- macOS only. Claude Meter is a native Swift app. It does not run on Windows or Linux. If you need cross-platform coverage, look at LLM-O-Meter, which covers both platforms, though it has fewer features.
- Requires Claude Code. The zero-config setup depends on Claude Code being installed and signed in. If you only use Claude via the web interface and never use Claude Code, this tool will not work out of the box. The OAuth fallback source requires a manual setup step.
- Anthropic does not publish exact limit numbers. The percentages Claude Meter shows are accurate, but the underlying absolute ceilings are not officially published by Anthropic and vary by plan. What this means practically: you know exactly what percentage of your limit you have consumed; you do not know what that represents in raw token count.
- macOS 14 or later. Older Macs running Ventura or below cannot run this app.
- Community project, not Anthropic. Claude Meter is independent. Jewei Mak maintains it and is responsive to issues on GitHub, but it is not an official Anthropic product. If Anthropic changes the API endpoint it relies on, the app may need an update before it works again.
Is This Worth Installing?
If you use Claude Code daily and you have ever lost your flow to an unexpected limit hit, the answer is yes. The install takes two minutes, the app runs silently in your menu bar, and the information it gives you costs you nothing but a glance.
The per-model cap visibility is the specific thing that makes this more useful than just checking the percentage buried in Claude’s settings. That cap is the one most likely to catch you by surprise, especially on Max plans where the Opus or Sonnet weekly budget runs out while your overall percentage looks healthy. Having that number visible means you can pace yourself before the wall, not after.
The widget is a nice secondary touch. If you keep a second monitor or a Notification Center sidebar visible, you can drop Claude Meter there and have an always-on readout without even needing to click anything.
The only scenario where this is not worth the install is if you are on the free plan or rarely use Claude Code. The free tier’s limits are more restrictive and the per-model caps are less complex, so the detailed breakdown adds less value. For paying subscribers who use Claude Code as a real part of their workflow, it earns its place in the menu bar.
Quick Reference: What Each Number Means
| Metric | What it tracks | When it resets | Who it affects |
|---|---|---|---|
| 5-Hour Session % | Usage in the rolling 300-minute window | Rolling, 5 hours after first use | All paid plans |
| Weekly All-Models % | Total usage across all models in 7 days | Once a week, fixed day | All paid plans |
| Per-Model Cap % | Usage of a specific model (e.g., Opus, Sonnet) | Weekly, same reset as above | Max plans primarily |
If you run Claude Code and you have hit a surprise limit at least once, Claude Meter is two minutes to install and free forever. Share this with anyone on your team who codes with Claude.
Sources
- Claude Meter GitHub Repository (Jewei Mak)
- Meet Claude Meter: Know Your Claude Usage at a Glance (jewei.net, June 24, 2026)
- Claude Code Documentation (Anthropic)
- Claude Code Limits: 5-Hour Caps Doubled (Verdent, May 2026)
- Claude Code Limits, Explained (CC Club, July 2026)
- Claude Code Usage Limits (DataCamp)
Leave a comment