Bondi is a system monitor for the Mac. It is in alpha. The job is one sentence: why this Mac is slow, with the number checked before you read it.
Activity Monitor is honest and useless at a glance. A browser is dozens of helper processes, so no single row looks like the problem. Bondi adds each helper to the app that started it. On one real reading that was 966 processes and 39 apps. Same Mac, both lists.
The sentence is the part I care about. Bondi measures every figure in code, from the Mac’s own readings. It decides what the sentence is about. On a Mac with Apple Intelligence it asks Apple’s on-device model to write the words around those facts. Other Macs can download a small open model from Settings. Nothing downloads unless you ask. Either way the model does not get to invent a number. If the tone, the subject, or any figure is wrong, Bondi throws the sentence out and uses its own.
History stays in a file on the Mac. The model runs on the Mac. Bondi contacts a server to check a trial or a license and to look for updates. That is the list.
A few things that are already in the alpha, because I use them: thirty days of history you can ask about in plain English, dev servers grouped by project, Docker containers, local models, a note on 125 common macOS processes, a separate volume per app, and a fan boost that is never slower than the fans already were. With its windows closed it sits around 0.3% of one CPU core. It asks before it quits, stops, or unloads anything.
The dates and the price are the ones printed on trybondi.app as of November 1, 2026. Alpha now. Beta next. Full launch is Tuesday, February 2, 2027, 8 am Pacific. A 14-day trial, no card, then a one-time purchase for one Mac. The first 250 licenses are $9.99. After that, $24.99. Students can email support from a .edu address. There is no subscription, and the on-device model is included.
I will correct this page when those figures change. The product page is the one to trust if we ever disagree.