Drifty

App-level time tracking tells you how long you had Slack open. It does not tell you whether you were in a deep thread resolving a production incident or refreshing out of habit. Drifty addresses that gap by capturing the active app, site, and session title in three-minute blocks — automatically, without timers — and running those signals through an AI classification engine that labels each block as Focus, Neutral, or Drift. The result is a daily timeline and weekly calendar view that reflects what the time was actually spent on, not just which windows were open.

The classification engine is the core technical bet. The vendor describes a ‘Context Engine’ that routes activity signals through fast model routes — DeepSeek V4 Flash, Kimi K2.6, and Qwen 3.6 27B, depending on the mode selected — applying profile rules to produce labels without asking the user to tag anything. Critically, the vendor states that only ‘minimal context label hints’ leave the Mac when using cloud routes, and no screenshots or screen recordings are taken in any mode. Users who want nothing leaving the device can run classification locally using a downloaded model that the vendor states consumes approximately 5–7 GB of RAM; that local route is free.

Drifty fits a specific profile well: a developer or freelancer on a Mac who wants a passive record of where focus actually went across the week, with enough detail to spot recurring drift patterns without maintaining any logging discipline. It breaks as a professional billing tool — there is no API, no export pathway described on the vendor page, no integrations with invoicing or project management systems, and no web or Windows access. Teams that need to reconcile focus data with client deliverables or share reports externally will exhaust what Drifty provides quickly.

All functionality runs on Mac only, and there is no self-hosted server option — the local AI mode keeps classification on-device but the application itself is a Mac download. The drift alarm feature is described as a nudge mechanism that fires during drift periods, giving the tool a light behavioral layer on top of the passive tracking and review workflow.