ClinicOS
Run the whole clinic from one calm place.
An operating system for a single clinic. The booking a patient makes at midnight and the chart a clinician signs the next morning run on one record — front desk, treatment room and books included. Deployed as a fresh instance on the clinic's own Vercel and Supabase accounts, so the clinic owns its data, its accounts and its bills rather than renting a seat in someone else's cloud.
- workspaces
- 5workspaces
- instance per clinic
- 1instance per clinic
- attachable modules
- 13attachable modules
The problem
A clinic runs on a booking widget, a spreadsheet, a card terminal and a folder of paper consents. Each one is defensible on its own. Together they mean the book, the chart and the ledger disagree, and on the day it matters nobody can say which one is right.
The usual fix is a large shared platform where the clinic is a tenant: their data in someone else's database, their bill growing per seat, and an export button that produces something nobody can import.
What we built
ClinicOS is one system for the front desk, the treatment room and the books — deployed as a fresh instance on the clinic's own Vercel and Supabase accounts. The clinic owns the data, the accounts and the bills, and there is no shared cloud to be a tenant of.
Five workspaces — front desk, clinical, money, growth and admin — write to the same record. There are no sync jobs and no export-import, so the book, the chart and the ledger cannot drift apart.
The parts that had to be exact
A held slot shows its real expiry rather than a hopeful one, so a deposit that is about to lapse says so. Charts become immutable the moment they are signed, each carrying a hash that proves the record has not moved since. The till is gapless and tamper-evident, which is the difference between a drawer that balances and a drawer you can defend.
Every booking, refund and AI call lands in an audit trail that records who did it, when, and under which permission. The assistant answers, books and takes deposits, and hands to a human the moment someone asks for one — AI-powered without being AI-dependent.
A closer look
What it actually does.
The patient side is the clinic's own front door
Not a booking widget bolted onto a website. The clinic puts its own domain and brand on a site that shows real availability, holds a slot honestly while the patient decides, takes the deposit, and carries the account, forms, receipts and memberships afterwards. A chat assistant books alongside it and hands over the moment somebody asks for a person.
One console, five workspaces, one record
The desk never has to leave the screen: the day's book with live holds, arrivals and walk-ins, the till, and an inbox of the few things that genuinely need a human — a deposit about to lapse, a caller the assistant handed off, a refund that was approved and logged. Behind it sits the audit trail, recording every booking, refund and AI call against a person and a permission.
The build is the clinic's, not a tier they were sold
Modules attach and detach at runtime, so a physio practice is not carrying a medspa's inventory module and a single-room clinic is not paying for multi-location. Because the instance runs on the clinic's own Vercel and Supabase accounts, the region, the bills and the backups belong to the clinic — and so does the exit, which is the part nobody asks about until they need it.
What is in it
The parts that do the work.
- Five workspaces — front desk, clinical, money, growth, admin — writing to one source of truth, so the book, the chart and the ledger cannot disagree
- Modules attach and detach at runtime; a clinic runs only what it uses
- Real-time booking with honest holds: a held slot shows its real expiry, with deposits and a patient portal behind it
- Charts immutable once signed, each carrying a hash that proves it, with consent capture built in
- A gapless, tamper-evident till and register
- An audit trail over every booking, refund and AI call — who, when, and under which permission
- HIPAA and GDPR program built in; AI-powered without being AI-dependent
- Python/FastAPI on Vercel Fluid Compute with React front ends, Postgres on Supabase, and a LiveKit voice agent that shares the same booking code rather than calling it over HTTP
Built on
- Python / FastAPI on Vercel Fluid Compute
- React front ends — staff console and patient portal
- Postgres on Supabase, one project per clinic
- LiveKit voice agent sharing the booking services directly
- Modular monolith with machine-enforced boundaries
- HIPAA and GDPR program built in