WorkMirror is a browser extension that watches your real workflow for a few hours and recommends AI tools based on what you actually do.
Ex: “You spent 2h 10m coding in Replit and switching files 74 times. Windsurf could cut context switching and save ~35 min/day.”
It’s privacy-first: no accounts & it's free.
The real challenge has been modeling workflow patterns reliably from noisy browser data. I’m using event timing + URL classification + interaction frequency to infer “work states” (coding, research, writing, idle) and feed those into a simple scoring model that ranks AI tools by expected time savings. Still iterating on the feature weighting and recommendation logic.
I'd love any feedback!