Healthcare SaaS and EMR

The desk books visits, group visits too, and quotes the travel fee up front. Each visit is posted to the nurses who can take it, and a nurse claims it from the app inside a set window; a nurse the client asked for gets the offer first.

The board shows each visit as late, overdue or still waiting on its exam. A nurse whose credential has expired, or has no expiry on file, cannot claim.

Nurses claim visits, follow the route and chart from a phone app. A chart keeps working with no signal and syncs later, never applying a change more than once.

The app reads a nurse's location only at the moments the nurse taps, never in the background.

Where a service needs an order, the order rests on a signed good-faith exam and is signed by a nurse practitioner, physician assistant or physician inside the exam's window. The database repeats each check, and a revoked order stops the next dose.

Every dose of a lotted product records its lot. An expired or recalled lot is refused by name, and a recall lists the patients, dates and nurses it reached.

Charts go to a provider for co-signature, and the medical director's protocols sit beside them.

Each practice gets its own AI assistant for calls and texts, on a model Telnyx hosts. It says first that it is automated, checks the address and its travel fee, quotes and books, and puts the caller through to a person when asked.

It gives no medical advice, and an emergency gets a fixed script, read word for word. Staff can try the assistant on invented patients in a playground.

An MCP connector lets Claude read the schedule, the dispatch board and the exam queue, and book, move or cancel a visit, through the product's own API.

A single TypeScript codebase on Postgres, with row-level security forced on every practice's tables.

• A Hono API and a worker, the only parts that open the database
• A Next.js web app for the desk, the providers and the medical director
• An Expo app for the nurses
• Shared zod contracts, each list of allowed values checked against its database column
• An audit trail on every practice's tables, and a test that probes every route with another practice's ids
• Every outside service behind a seam with a test driver, so the whole product runs on synthetic data