A complete SaaS product for general aviation, built solo.

A flying club or a group of co-owners usually keeps its administration in a shared spreadsheet and a chat group. That works until it does not: two people book the same aircraft, the hours do not match the previous flight, and at the end of the quarter someone has to work out by hand who owes what.

They are small organisations with a real accounting problem: every flight is a transaction, every aircraft has its own rate, and the maintenance interval hangs on a meter reading that changes with each flight.

Two-click booking, a flight log attached to the booking, automated invoicing based on Hobbs or Tacho time, a ledger per pilot, maintenance forecasts per aircraft, and a subscription model with iDEAL and SEPA direct debit.

On top of that an MCP server, so an AI assistant can book directly, check a balance or estimate what a flight will cost. That was not a marketing experiment: pilots already plan their flight in conversation with an assistant, and the booking is the logical next step.

NestJS and PostgreSQL on the back end, React on the front, Stripe for payments, multi-tenant throughout. Built with AI agents under our own review, with a test suite as the gate.

The hard parts were not the screens but the money: a balance that reconciles with the customer record, an expired subscription that falls back cleanly instead of hanging, a webhook that remembers whose document it already processed. Things like that go wrong once a year and then immediately at a real customer, so that is where most of the tests are.

That a small team with AI agents can stand up and keep running a complete, paid SaaS product — including payments, invoicing and support. Not that AI takes over the work: every decision about the data model, every trade-off about what stays out, and every review remained human.

And above all: it is why we can talk about AI-driven delivery without a slide. We know what it costs, because we paid the bill.