Focus Tracker – Context Switch Analyzer

Track distracting tab switches and improve your focus. All data stays on your device.

Focus Tracker measures what your tab switching actually costs you. Most trackers count minutes per site. That number tells you little — an hour on YouTube is research for one person and a lost afternoon for another. Focus Tracker measures something harder: whether a switch broke your work, and how long it took you to get back. WHAT IT MEASURES • Real context switches, not every tab change. Moving from your docs to your editor is one task continuing. Moving from your editor to social media is not. Only the second is counted. • Active time only. The clock stops when the window loses focus, when you go idle, and when your machine sleeps. Time you weren't there is never counted against you. • Resumption cost. An interruption costs more than the minutes it took — there's the re-orientation afterwards. That overhead is estimated and added to the total. • Five named kinds of distraction: time on a distracting site, going back to the same site again and again, tab thrashing, ping-ponging between two sites, and breaking off work earlier than you normally do. Every number comes with the measurement behind it. If a minute was counted against you, the popup tells you why. CALIBRATED TO YOU There's no universal definition of a distraction, so the thresholds come from your own history — how long your glances usually last, how long your work blocks usually run. Three sensitivity levels adjust how strict it is. You can override any site's category, exclude domains entirely, and overrule any individual verdict. The focus score stays blank until there's enough tracked time to mean something, rather than showing you a flattering 100. WHAT YOU SEE Today: your score, what's being tracked right now, what today's distractions cost, and recent activity. Trends: a log of what you actually worked on, week over week, cost per site, focus rhythm by hour and by weekday, and your longest uninterrupted blocks. When your attention starts to wobble, you get a nudge and the option to declare a break. A break you choose is not a lapse — the clock stops and nothing is counted against you. OPTIONAL ON-DEVICE INTELLIGENCE Two models can run locally in your browser to sharpen the judgements, both off by default: • A small embedding model, included in the extension, that decides whether two pages belong to the same task — so a conference talk about what you're building counts as work, not leisure. No download, works offline. • A larger instruct model that works out what an unfamiliar site is for instead of relying on a built-in domain list. This one is a one-time download if you choose to enable it. With both off, the tracker still works using word overlap and its built-in categories. PRIVACY Everything stays in your browser. No account, no server, no sync, no analytics. Nothing about your browsing is ever transmitted. Page content reading is off by default. If you turn it on, it reads a short summary of each page — heading, description, first paragraph — to tell tasks apart more accurately than a title can. Only individual content words are stored; page text never is. The only network request this extension ever makes is downloading the optional large model, if you enable it.