Barn Gym · Personal training and group fitness, Burwash, East Sussex
Bookings, plans, meals and members on one platform.
“Workout and meal plans written by hand for every client.”
From the project docs


What we built
What we built
Barn Gym runs personal training and group fitness in Burwash, East Sussex. Before, it ran on several paid tools that did not talk to each other: a community platform in one place, booking in another. Consultations, coach diaries and time off were tracked outside any system. Every workout and meal plan was written by hand, for every client. Members had nowhere to log training, food, photos and measurements together, and there was no admin view of bookings, credits or how each coach was doing. Every one of those gaps was a job somebody did by hand.

Now members have one dashboard, with their next sessions, their coach and their streak on the first screen. They book classes, personal training and consultations, and join a waitlist when a session is full. They log workouts set by set, with personal records, and track their eating with a macro calculator for protein, carbs and fat, alongside measurements, check-ins, progress photos and streaks. New people book a consultation through a public page, fill in a health form first, and a coach approves the booking.
Coaches get their own portal. Their schedules and time off now live inside the system, where the owner can see them, and coach analytics finally show how each coach is doing. Plans are built from a library of 205 exercises and a database of 120 meals, so a coach assembles a plan from parts that already exist instead of writing it from scratch for every client. The hand-written plan, the line at the top of this page, is the job this part was built to retire.

The AI is in the meal planner, because a week of meals for every client is exactly the kind of job a machine should draft. A coach picks a client, presses Generate Full Week, and Google’s Gemini drafts seven days of meals for that client in seconds, with calories and macros for each one. The coach checks the plan, regenerates any day that is off, and decides what the member gets. The AI drafts. The coach signs off. What reaches the member is a plan a coach has put their name to, not a chatbot’s best guess.
The platform replaces the paid community tool and the separate booking. The community now lives inside it: spaces, channels, challenges, polls, events and coach-to-member messaging, with 37 of 37 community API tests passing - automated checks that each part does what it should. The admin side has dashboards, bulk actions, audit logs and CSV export, so the data can leave with its owner. Members, coaches and the owner now work from the same system, and the same numbers. It runs to about 50,000 lines of code across roughly 85 database tables, and each gym’s branding is set separately. That last part matters, because the system is now becoming a product for other gyms. The foundations for running more than one gym are in place, and the work to keep each gym’s data fully separate is under way.
The numbers
The numbers
What it runs on
Hono and TypeScript on Cloudflare, with D1 for data, R2 for files and Gemini 2.0 Flash for the meal plans.
What’s next
AllIn1Gym: the same system, becoming a product for other gyms.
Related: An app, yesterday · App development · AI automation · Web development
Next case studyPanasonic Northern Europe →
Free business review
Tell us what’s slow. We’ll tell you what to fix first.
A call of 20 to 30 minutes. We find the gaps, the leaks and the quick wins, and what each one is worth to you. Fixed price after that, no hourly rates. No 200-page deck.
Or tell us what you want to achieve and Dan will come back to you within one working day. Email dan@danielnicholls.com.