71% weekly quota burned by Monday on one Opus refactor. Local token estimators were off by 30%+. The metered billing split is why.Claude Meter Pro: a background agent that saves your usage. Watches your Claude Code and Claude.ai calls, finds savings without losing quality, and (with permission) ships the fixes for you. Join the waitlist.
macOS menu barLive Claude Pro & Max usageFree & open source

Did you get?

See exactly how much of your Claude Pro/Max plan is left before you blow past it: the rolling 5-hour, weekly, and pay-as-you-go numbers, sitting next to your clock. Same numbers claude.ai/settings/usage shows, to the integer. One brew command; the browser extension forwards your existing claude.ai session, so no DevTools and no cookie paste.

Free, open source, MIT licensed. No account or card: we email the .dmg plus one brew command, and your menu bar lights up within a minute.

Server-truth quota5h + 7-day windows60s refresh
~ / claude-code
Live preview · matches claude.ai/settings/usage to the integer
01Built for

Built for the people Anthropic’s rolling quota change hit hardest.

Claude Max developers running agentic loops. Claude Pro heavy writers hitting the weekly wall midweek.

01 / Session

Live rolling 5-hour window

Watch the Pro/Max session quota tick down as you prompt, so you know when to stop before Claude cuts you off mid-refactor.

→ updates every 60 seconds
02 / Weekly

Weekly quota and reset clock

See how much of the 7-day allowance you have left and exactly when it rolls over, not a vague 'try again later' banner.

→ reset timestamp to the minute
$
03 / Spillover

Extra-usage balance

Track your pay-as-you-go spillover in dollars, so the metered billing Anthropic rolled out in 2026 stops being a surprise.

→ current billing cycle total
04 / Auth

Browser extension auto-auth

Install the Chrome/Arc/Brave extension and it forwards your existing claude.ai session. No DevTools, no manual cookie paste.

→ Chrome · Arc · Brave
05 / Privacy

Source on GitHub

Rust source on GitHub. usage-data network egress is claude.ai itself; anonymous health telemetry is opt-out.

→ auditable · written in Rust
$ claude
62% · $3.40
$ _
06 / CLI

CLI included

Pipe the same numbers into shell prompts, tmux status lines, or scripts. Useful next to ccusage, which tracks something different.

→ tmux · starship · zsh prompt
05The alarm

At 95%, your menu bar starts flashing red.

The instant your 5-hour rolling window crosses 95%, the entire ClaudeMeter title in the menu bar blinks red at 500 ms, macOS fires a system notification, and Sosumi rings three times. You finish your prompt before the wall hits, not after.

Muted by default. Real app plays Sosumi ×3 with a 120 ms gap. Visuals loop on autoplay.
  • Title flash cadence: 500 ms, RGB 215,58,73. Whole title flips together, not just the percent.
  • System notification: “Claude usage at 95%” with the 5-hour subtitle and a wrap-up prompt.
  • Sosumi ×3 via afplay, 120 ms gap. Fires once per 5-hour window; rolls over on reset.
02Comparison

ClaudeMeter vs ccusage.

They solve different problems. ccusage reads local Claude Code JSONL; ClaudeMeter reads the plan quota Anthropic actually enforces.

FeatureClaudeMeterccusage
Reads server plan quota (5h / weekly)YesNo
Reads local Claude Code token JSONLNo (by design)Yes
Browser extension auto-authYesN/A
macOS menu bar UIYesCLI only
LicenseMITMIT
Read the full comparison
03Why people install it

From the threads where people find out the hard way.

I typed 'test one two three' into Claude Code. That put me at 12%.

Max 20x subscriber at $200/month. Getting rate-limited on every Claude surface with only 18% session used.

I used it a little Friday, a little Saturday, maybe 10 minutes Monday, and I hit the weekly limit already.

04Install

Stop getting surprised by “message limit reached”.

ClaudeMeter reads the server-truth numbers Anthropic enforces for Pro and Max: the rolling 5-hour window, the weekly quota, and the extra-usage balance. No cookie pastes. No token guesswork.

View source on GitHub
5h62%Week41%$3.40