
See how much of every AI subscription you've used, in one place
One widget for Claude, ChatGPT and whatever else you pay for. It never invents a number: every reading is tagged live, manual or error.
- 0 dependencies
- Menu bar + widgets
- v0.5.3 on PyPI
Anuj Patel, Lead Product Designer
7+ years across healthcare, B2B and enterprise. I find the structure under messy domains — infinitely custom insurance policies turned out to be five money-flow patterns, a nine-step booking form turned out to be one — and build the systems that compose themselves from it.
See the workEach one left something behind: a pattern registry that renders any policy shape, one card component carrying fourteen states, a shared design system that held consistency across teams as the surface area grew.

Members couldn't tell what was covered, so they paid out of pocket and claimed it back. Sales can now sell any plan shape and the interface renders it — no redesign.
Measured in production, first six months after launch

Users expect to connect within fifteen minutes; a nine-step booking form became one. Three years on, it is still the journey the consultation business runs on.

A third of a doctor's day goes to documentation, often without a computer at all. Template reuse became how the product is used, not a power-user feature — template usage up 200%.

Medically-reviewed videos, surfaced where members already were.

One widget for Claude, ChatGPT and whatever else you pay for. It never invents a number: every reading is tagged live, manual or error.

Holds an agent's risky actions until a human decides, then turns the branch into a short narrated page for the reviewer.
I’m Anuj Patel, a Lead Product Designer with 7+ years across healthcare, B2B and enterprise products.
Most of my work arrives as someone else’s brief and doesn’t stay that way. The benefits brief asked for a nudge toward cashless; research showed members couldn’t tell what their plan covered, so there was no decision left to nudge — and the roadmap was rebuilt around comprehension instead.
I run the working sessions where that gets settled, with backend, frontend, ops and onboarding in one room. Things get killed there too: configurable per-service caps went once engineering showed the data model couldn’t enforce them, and we shipped an honest dual-constraint display instead of a promise. Lately, that includes how AI is changing both what we design and how.
I care about solving real user problems and making products genuinely better.Download resume