Token Time

Token Time is a macOS menu-bar app that keeps a running tally of how many AI tokens you burn through and what that's costing you, updated live. The framing the makers reach for is deliberately familiar, it's Screen Time but for your tokens. The idea is to take a number that normally hides inside a provider's billing dashboard and park it at the top of your screen, where you'll actually glance at it during the day. The problem it targets is quiet drift. When you spend a full day working alongside Claude or another model, tokens leave in a steady trickle rather than one visible chunk. A single long, looping session can cost more than an ordinary week, and most people don't notice until the invoice lands. Token Time pulls that feedback forward to the moment it's happening, so a heavy afternoon reads as a climbing meter instead of an end-of-month shock.

The core of it is the menu-bar meter, which shows a live token count and the current day's spend without you opening anything. Click into it and the usage is laid out by the hour across the day, which is exactly where the runaway sessions stand out. Instead of one flat total, you get a shape for the day, and a spike at two in the afternoon tells you more than a single figure ever could. It also breaks cost down by model and holds the current day up against your usual baseline. That split matters when you juggle a cheap model for drafts and an expensive one for the hard problems, because it shows you where the money is actually going rather than lumping everything together. The baseline comparison is the part that flags trouble early, since a day that runs well above your normal pattern is usually a sign that something looped or an agent went off on its own.

That hourly and per-model view earns its keep once you start running agents or background jobs that keep working while your attention is elsewhere. An agent that retries, re-reads a long context, or loops on a task can quietly move a lot of tokens in a short window, and a flat monthly total would never show you which hour it happened in. Watching the shape of the day turns that from a mystery into something you can trace back to a specific run, so you can change how you set it up the next time instead of guessing.

For the stretches where you get absorbed and stop checking, there are full-screen reminders. When you cross a threshold you've set, a reminder briefly takes over the screen and then clears itself after ten seconds, so it interrupts just enough to register without derailing your work. You decide how often those nudges fire, tuned in steps of a few million tokens, which lets a light user and someone running heavy batch jobs both settle on a rhythm that doesn't grate. The reminders are a nudge rather than a hard cap, and that distinction is deliberate. Token Time doesn't cut you off or throttle anything, it simply makes sure you know where you stand, so the choice to keep going or ease off stays with you. Because you tune how often it fires in steps of a few million tokens, the same app can sit quietly in the background for someone who mostly chats and speak up often for someone pushing large jobs through all day. It's a meter and a gentle alarm, not a gatekeeper.

Privacy isn't an afterthought here. There's no account to create, nothing syncs to a cloud service, and the app sends no telemetry, so the record of how you use AI never leaves your machine. For anyone who'd rather not hand a running log of their model usage to yet another third party, that local-only posture is a genuine part of the appeal, and it means the app behaves the same whether or not you happen to be online.

The audience is developers and AI power users, the people who keep a chat or an agent open next to the editor and quietly rack up tokens all day. If you've ever reached the end of a billing cycle with no clear idea where the spend actually went, this is built to answer that question while you can still do something about it, rather than weeks later once the money is already gone. It fits neatly into a workflow where you're moving fast and don't want to keep tabbing over to a usage page to check yourself.

What sets it apart from a provider's own dashboard is immediacy and independence. It watches usage across your work instead of living behind a login you have to remember to open, and it does that without collecting anything about you. Access is a one-time purchase rather than a subscription, which is worth noting on its own since so many small utilities lean on recurring fees. It's listed at six dollars at launch against a regular price of eight, so once you own it there's nothing to renew. It runs on macOS 13 and later on Apple Silicon, and the makers publish a support address for questions.