Usage4Claude: Claude and Codex quota limits in the macOS menu bar
Monitor all your Claude/Codex usage limits in real-time from your macOS menu bar - supports 5-hour, 7-day, extra usage, 7-day Opus and 7-day Sonnet quotas.
At a glance
- What is it?
- Usage4Claude is a native SwiftUI menu bar app that reads your Claude subscription quota (5-hour, 7-day, extra usage, per-model weekly) and optionally Codex limits. It is a read-only display layer, not a usage controller, and its accuracy depends entirely on the session key it extracts at login.
- Who is it for?
- Adopt Usage4Claude if you run Claude Code or the Claude apps on macOS 13 or later and keep hitting a 5-hour or 7-day wall without warning. Skip it if you are not on macOS, if you will not hand a session key to a third-party app, or if you need the tool to act on quota rather than display it.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 9 days ago.
- What is it written in?
- Mainly Swift, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The quota wall problem Usage4Claude addresses
Claude subscriptions meter usage across several independent windows: a 5-hour limit, a 7-day limit, extra usage, and weekly per-model limits for models such as Opus or Sonnet. The README states that all Claude surfaces (Claude.ai, Claude Code, the desktop app, mobile, and Cowork) draw on the same pool. That is convenient for billing and inconvenient for planning, because the number you are about to exhaust is not visible from inside the editor where you are exhausting it.
Usage4Claude puts those numbers in the macOS menu bar as rings. The intended user is a developer or heavy Claude user on macOS who wants a glanceable reading rather than a dashboard tab. The README also covers Codex, so a user paying for both can watch both in one place, in a dual-column view when a Codex account is added.
How the menu bar rings get their numbers
The app is a native SwiftUI menu bar application, built from Usage4Claude.xcodeproj with Package.swift present for the Swift package side. Credentials live in the Keychain: the README says the Claude Session Key and the Codex authentication token are stored there and never in plain text, with App Sandbox enabled. Data is kept locally, and the README states the app does not collect or upload personal information.
Login is handled in a built-in browser. For Claude, the README says the session key is extracted automatically after login; for Codex, the same built-in browser handles ChatGPT authentication. That session key is the whole data path. The app is a client of the same usage information the web interface shows, which is why the README can claim coverage across web, CLI, desktop, mobile and Cowork: they share the quota, so one credential reads all of it.
Refresh is either a fixed interval of 1, 3, 5 or 10 minutes or a four-level adaptive scheme, with a manual refresh button that has a 10-second debounce. Notifications fire at 90% usage and when a quota resets. A smart display mode shows whichever limit types have data; a custom mode lets you pick the combination yourself.
Installing Usage4Claude and reading your first quota ring
The README points at the releases page, and the repository ships appcast.xml, which is the Sparkle-style update feed a menu bar app uses to check for new versions. The build targets macOS 13.0 or later and Swift 5.0 or later. If you would rather build from source than download a release, the project is an Xcode project with a Swift package manifest alongside it:
git clone https://github.com/f-is-h/Usage4Claude.git
cd Usage4Claude
open Usage4Claude.xcodeprojBuilding in Xcode and running the target produces the menu bar item. On first launch the README describes a setup wizard; the Claude login happens in the built-in browser and the session key is then stored in the Keychain. After that, the menu bar shows a ring per limit type. Clicking the icon opens the detail view; right-click is the second access path the README mentions.
The keyboard shortcuts the README lists are worth memorizing, since they are the whole interaction surface:
⌘R manual refresh
⌘, settings
⌘Q quitSettings is where you choose percentage-only, icon-only, or icon plus percentage, set the refresh interval, and switch appearance between system, light and dark. If you add a Codex account, the display can switch to a dual-column layout; without one, the README says the Claude-only experience is unchanged.
Where Usage4Claude breaks or misleads
The dependency on an extracted session key is the sharpest edge. v3.4.0 is titled "Icon Size Setting & Account Credential Fixes", which tells you credential handling has needed repair in the recent past. If the extraction path fails after a Claude-side login change, the rings go stale or empty, and the app has no documented way to fall back to an official API, because there is no official quota API described in the README. It reads what the web session can see.
Platform lock-in is total. This is a macOS 13+ SwiftUI app. There is no Windows or Linux build in the repository layout, and the menu bar is the product. If your team is mixed-platform, this tool covers only the Mac users.
The Codex reset badge is explicitly a beta and explicitly third-party: the README says the data comes from codex-reset.com, described as an independent community project rather than an official API, and that most of the time there is nothing to show by design. Treat that badge as a rumour feed, not a fact about your account. It can be turned off in Settings.
Finally, the app displays quota; it does not manage it. There is no documented way to pause a job, throttle a model, or block a request at 90%. The 90% notification is a warning, and the manual refresh has a 10-second debounce, so during a fast burn you are watching a number move, not stopping it.
Usage4Claude against the other menu bar meters
The related searches around this project name CodexBar, SessionWatcher, ClaudeMeter and Clawdmeter, so the category is crowded with small menu bar utilities. The distinguishing choice in Usage4Claude is breadth of limit types combined with Codex in the same app: the README lists 5-hour, 7-day, extra usage, and weekly per-model limits for any number of models, plus Codex 5-hour, 7-day and credits, in one menu bar item with an optional dual-column view.
A narrower tool that only shows a single 5-hour percentage is easier to reason about and has less to break. Usage4Claude's smart display mode exists precisely because the field set varies by plan and account, which means the app carries parsing logic for limit types that some users will never see. If you only care about one number, that complexity buys you nothing. If you juggle a Team or Max plan with per-model weekly caps, the extra fields are the reason to pick this over a simpler meter.
Maintenance, releases and the MIT licence
The repository is not archived, and the last push was on 2026-09-07, days after the v3.4.1 release on 2026-09-04. The release cadence visible in the release history is uneven: v3.3.0 landed on 2026-07-14 with German localization and multi-model weekly usage, then v3.4.0 and v3.4.1 arrived on the same day in September. That pattern suggests bursts of work around credential and display fixes rather than a steady drip, which matters if you are pinning a version for a team.
Upgrade cost is low by design. The app ships appcast.xml, so update checks and the menu bar badge plus rainbow animation described in the README are built in; there is no migration step documented for settings or Keychain entries between v3.4.0 and v3.4.1. The licence is MIT, which permits commercial use, modification and redistribution provided the copyright notice and permission notice are included. That is the extent of what the repository states; it is not legal advice, and if you redistribute a modified build you should read LICENSE yourself.
One gap worth naming: the README does not document rollback. If a new version misreads your plan's limit fields, there is no downgrade procedure described beyond fetching an older release.
Editorial conclusion
Adopt Usage4Claude if you run Claude Code or the Claude apps on macOS 13 or later and keep hitting a 5-hour or 7-day wall without warning. Skip it if you are not on macOS, if you will not hand a session key to a third-party app, or if you need the tool to act on quota rather than display it. Before relying on it, verify that your plan exposes the extra usage and per-model weekly fields the app expects, and confirm your session key survives a re-login, because the credential extraction path is where v3.4.0 spent its fixes.
Frequently asked questions
What is Usage4Claude and what does it monitor?
It is a native macOS menu bar app that displays Claude subscription usage quota, and optionally Codex quota. The README lists Claude 5-hour, 7-day and extra usage limits plus weekly per-model usage, and Codex 5-hour, 7-day and extra usage or credits.
How to tell usage on Claude with this app?
After logging in through the built-in browser, the menu bar shows a ring per limit type, with percentage-only, icon-only, or icon plus percentage display modes. Clicking the icon opens the detail view, and the README also mentions a right-click access path.
How much is usage in Claude, and does Usage4Claude show it?
The app does not define the limits; it reads the quota your account already has. The README states it supports Free, Pro, Team and Max subscriptions and displays whichever limit types have available data.
What macOS version does Usage4Claude need?
The README badges specify macOS 13.0 or later and Swift 5.0 or later. It is a SwiftUI app, so there is no Windows or Linux build in the repository layout.
Does Usage4Claude store my Claude credentials safely?
The README says the Claude Session Key and Codex authentication token are kept in the Keychain rather than in plain text, with App Sandbox enabled and all data stored locally. It also states the app does not collect or upload personal information.
Community notes