Human Token tracks your daily cognitive load as weighted tokens — mirroring how LLMs process context.
It sits in your menu bar, watches which apps you use, and turns that into three numbers that tell you how your day is going — mentally.
Every app you use gets counted — coding, writing, meetings, browsing, AI tools. No setup. It just works in the background from the moment you open it.
Shows how varied your work is today. Stuck in one app all day? Low temperature. Mixing coding, writing and research? Higher. The sweet spot is somewhere in the middle.
Think of it like a fuel gauge for your brain. The average person can do about 4 hours of real deep work per day. Saturation tells you how much of that you've already used.
Walk away for 10+ minutes and your token count goes down. Real rest genuinely restores your mental capacity — Human Token tracks that too, not just the work.
The live velocity chart shows your token rate hour by hour. You'll quickly spot when you do your best work — and stop scheduling meetings there.
Every day is saved on your Mac. Come back a month later and see exactly how you've been spending your mental energy — no account, no subscription, no cloud.
Human Token weights different cognitive activities differently — just like transformer attention weights some tokens more than others.
Human Token detects your current cognitive mode from the running app stack — no setup required.
IDE + Terminal running
AI tool in foreground
Zoom / FaceTime active
Standard activity
8–14 app switches/hour
15+ app switches/hour
Each metric surfaces a different dimension of your cognitive state — in real time, from your Mac.
How varied is your thinking today? Low means you've been stuck in one mode all day. High means you've mixed different types of work. Both extremes are fine sometimes — this just tells you where you are.
How much of your daily mental budget have you spent? Research consistently shows the brain can sustain about 4 hours of real deep work per day. 100% means that tank is empty.
Rest genuinely frees cognitive context. Wiehler et al. (2022) showed brief pauses don't reduce prefrontal glutamate — recovery requires at least 10 minutes of real disengagement.
| Break | Recovery |
|---|---|
| < 10 min | None |
| 10–30 min | Up to 600 tokens |
| 30–120 min | Up to 1,800 tokens |
| > 2 hours | End of day |
No account, no cloud, no telemetry. Here's exactly what the app reads — and what it never touches.
Reads only the domain of your active tab every 30 seconds — to detect AI tools and web meetings. Never page content, never other tabs. You can decline and the app still works.
Reads which app is in the foreground — only its name and bundle ID. Used to classify your cognitive mode and token type. Nothing else.
Uses Spotlight to detect code saves in the last 45 seconds. Only reads the file extension (e.g. .swift) — never file contents.
Checks seconds since the last keyboard or mouse event. No keystrokes captured, no clicks recorded. Used only to detect breaks.
Runs git diff --shortstat to count lines changed. Never reads commit messages or file contents.
No microphone, camera, contacts, or screen. No keylogger. No internet calls — ever. All data lives only in your Mac's local storage.
No account. No setup. Just open and it works.
Download Human Token