Show HN: A macOS menu bar app showing Claude subscription usage

What happened

I'll look up the Show HN thread so the paragraphs stay grounded in the actual product and numbers, then write the body as plain prose.The product name is usagent; I’ll pull the original Show HN and any repo notes so the write-up stays on those facts.The summary is truncated, so I’m checking the local ledger and any staged writer output for the full Show HN text.The pipeline summary is cut off. I’ll check the GitHub source for the rest of the author’s description, then stay inside those facts.A Show HN post on the HN Claude/Codex/Fable feed introduces usagent, a macOS menu bar app that keeps Claude subscription usage limits on screen instead of behind a hunt. The author uses multiple Claude products and said they constantly check those limits, but there was never an instant way to do it. Their existing loop was to open claude.ai, go to settings, then usage, or to stop in the CLI and hit /usage. Both of those paths already work. Neither stays visible while they are in another product. They wanted the same information always visible and at reach, so they built usagent. The app sits in the menu bar and always shows usage limits, including the 5h window and the weekly window that already sit on a Claude subscription. That is the event: a multi-product Claude subscriber who was polling official surfaces by hand shipped a glanceable client for the two clocks they already had.

The mechanics are a status surface, not a new quota system. usagent is a macOS menu bar app that always shows usage limits and tracks both the 5h window and the weekly window. By default it does not try to render both at once. It shows whichever window is closer to its limit, so the bar is a binding-constraint indicator rather than a full report. A click is the only extra gesture described: you see both windows with full info. That split is the product. The 5h window and the weekly window fail in different ways on the same plan. One can sit far from its cap while the other is the reason the next long session dies. Defaulting to the closer window treats menu bar space as scarce and treats the click as the place for the complete pair. The author does not claim a third clock, a new plan type, or a replacement for every field on the claude.ai usage page. They claim two windows, always on, with the tighter one in chrome until you ask for both.

The technical detail

For engineers and builders who hop across Claude products, the tax is not that usage is secret. The tax is that usage is one or two navigations away every time a decision depends on it. Opening claude.ai, then settings, then usage, or dropping /usage into the CLI, is a context switch in the middle of a run you are trying to finish before a window resets. Multiple Claude products on one subscription draw from the same 5h window and the same weekly window, but the official checkers live inside a single product at a time. You can burn the shared budget in one surface and only learn it when another surface refuses the next job. A menu bar reading is useful because it answers the operational question without leaving the editor, the browser, or the terminal you are actually in: which clock is closer to its limit, and do you still have room to start the next job. That is a scheduling input, not a vanity metric. It changes whether you open a heavy agent loop, switch to a lighter product, or wait for a window to move.

Why it matters for builders

The competitive context is the first-party pair the author already used. claude.ai settings usage is the web view. /usage in the CLI is the terminal view. usagent does not add a third plan and does not invent a new limit model. It relocates the existing two windows into macOS chrome so they are always visible. That is a small market and a real one: people who already pay for Claude, already bounce across multiple Claude products, and already know they have a 5h window and a weekly window, yet still have to go looking. Official surfaces optimize for completeness when you seek them out. A menu bar app optimizes for the moment you do not want to seek them out. The comparison that matters is not dashboard versus dashboard. It is interrupt-driven checking versus a default that surfaces the tighter of the two clocks. If the official pages remain the source of truth, usagent is competing with the habit of opening them, not with Claude itself.

Market and competitive context

The practical test is whether the default closer-to-limit rule is the number you would have gone to settings or /usage to find. If the 5h window is the one that usually stops you, the bar should spend most of its time on that window. If the weekly window is the one that sneaks up after several 5h cycles, you will learn quickly whether a single collapsed reading is enough or whether you still click for both with full info before you start a heavy day. Watch whether you stop opening claude.ai settings usage as a reflex, and whether /usage remains the place you go when the menu bar is not enough. The Show HN only promises always-visible limits, the two windows, a closer-to-limit default, and a click for full info. Use that as the acceptance bar, not a broader usage console. If you still need the CLI after a week, the menu bar solved glanceability and did not solve diagnosis.

What to watch next

Open questions sit next to that acceptance bar. The summary does not say how usagent reads the 5h window or the weekly window, how fresh the numbers are, which of the author's multiple Claude products it can see, or what full info includes once you click. An always-on menu bar client next to a paid Claude subscription is a credentials and trust question as much as a UI question, and that side is unspecified here. There is also a product risk in the default itself. Showing whichever window is closer to its limit is good glance design and a way to hide the other clock until a click. If you need both windows to decide whether to wait, switch products, or keep going, the collapsed view can be incomplete at the exact moment it looks decisive. The source writeup also cuts off as the author starts to say what they personally do with the app, so anything beyond the two windows, the default, and the click is not in the record. Treat those missing pieces as things to verify before you leave usage for a paid plan in a third-party menu extra.