Live · Telegram app + Web · 12 languages
A platform for a trainer and their clients.
Anvil Discipline - a Telegram app and Web for trainers and their clients. An AI trainer and an AI nutritionist, an 873-exercise catalogue, nutrition by photo and voice, ready-made diet protocols with an auto-assembled grocery basket, a women's cycle module with a «cycle × training × nutrition» cross-link, booking with no double bookings, payment via Telegram Stars. 12 languages.
- The problem
A trainer runs clients in Telegram chats - and loses half the revenue to it.
Workout plans - in voice messages. Booking - by forwarding calendar screenshots. Payment - bank transfers. Reminders - on a verbal promise. One trainer physically can't handle more than 5-8 clients in parallel. Growth tools aren't on the menu.
Pain 01
Programs - in head and voice notes
The plan goes to the client by voice or screen in a chat. A week later the client doesn't remember exactly what to do. Progress isn't visible, drop-offs are frequent.
Pain 02
Nutrition - a separate world nobody tracks
The trainer sets «-200 kcal, +30g protein». The client tries to count in a third-party app, gives up in a week. Without nutrition, workouts deliver half the result - and the trainer sees it after the fact.
Pain 03
«What should I eat?» - a question with no answer
The client doesn't need calorie counting - they need a ready system: what to buy, what to cook, how not to break. The trainer isn't a nutritionist, and writing day-by-day menus for everyone is unrealistic. The client googles diets and makes it worse.
Pain 04
Booking and payment - by hand
«When works on Tuesday?» - the client replies an hour later, the trainer is already busy. Double bookings, cancellations. Payment - bank transfers: sometimes forgotten, sometimes «next week». Excel or lost money.
Pain 05
Reminders - on the trainer
«Hi, 18:00 today» - the trainer writes to 10 clients every morning. Spends an hour, forgets one, loses a session. With each new client the load grows linearly.
One product where the trainer runs the entire client relationship.
Workouts, nutrition, protocols, schedule, dialogues and payment - in one system. The client lives in Telegram, the trainer in their cockpit, and both see one picture.
AI trainer - a plan draft for the goal
The trainer sets the goal, days per week, focus, session length and limitations («knee hurts») - the AI trainer assembles a draft from the catalogue, automatically excluding contraindicated exercises. The trainer edits and activates: the human always has the last word. A solo client without a trainer gets the same AI trainer.
Plan editor with an 873-exercise catalogue
A catalogue with GIF demos, custom exercises via quick or full entry, with a preview before adding. RPE is opt-in, for the advanced. The client ticks exercises off right in the Telegram reminder.
AI nutritionist
Analyses the client's diet and answers questions in an «ask the nutritionist» format: what about protein, what to substitute, what to drop. Not generic internet advice - a review of the client's actual food log.
Nutrition - by photo or by voice
The client photographs a meal - the app recognises composition and counts macros. Or dictates by voice, scans a barcode, photographs a label, uses common measures - «a glass», «a spoon» - or repeats yesterday with one button.
Diet protocols - with a grocery basket
Ready-made nutrition systems by days and meals: anti-inflammatory, Mediterranean, DASH, 5:2, high-protein and more. The client enrols - and the grocery basket for the chosen day range assembles itself: «4 pcs», «1 bunch», «2.5 l», by category. Meals get ticked off, adherence is visible.
Women's cycle module - cross-linked with training
A phase ring, a 3-12-month forecast, a sensation log. Data isn't shared with the trainer by default. The unique part: workouts, nutrition and weight break down by cycle phase - first an honest picture from her own data, optionally an AI observation with no prescriptions.
Voice in the trainer ↔ client dialogue
Hold - dictate - confirm the transcript - send. The other side gets the message as a Telegram push, even with the app closed. A skipped-workout reason can be dictated by voice too - the trainer gets it as text.
The trainer's day - on one screen
Clients in order of importance: declines and overdue first, then snoozed and pending. Per client - progress, the day's nutrition, the decline reason as a quote, unread messages.
Booking and payment in Stars
The trainer sets slots - the client picks; a double booking is technically impossible. «Anvil Trainer» tiers price by client count, and a trainer's invited clients pay nothing. Payment in Telegram Stars right in the chat; ₽ cards in progress.
Reminders and motivation - by themselves
A reminder at the set time in the client's timezone - with per-exercise checkboxes. Snooze once; «can't make it» - with a reason. The trainer sees every skip immediately; a quiet client gets a gentle AI ping, at most once in 4 days.
The engineering that holds it.
On the outside - a simple «tap and it works» product. Inside - the mechanics without which products like this fall apart at the second dozen clients.
/ 01
One backend - two faces
The app and the Telegram bot run on one database: the client ticks an exercise in the chat - the trainer sees it in the cockpit instantly. No «second version of the truth».
/ 02
10 background services
Reminders in each client's timezone, skip tracking, AI pings to the quiet ones, water, weekly digests, a Monday report to the trainer, journal broadcast. Everything that «should happen by itself» - happens by itself.
/ 03
Guards at the database level
Double bookings blocked at the DB level, the trial protected from re-farming, AI-credit debits atomic. Edge cases closed where they can't be bypassed.
/ 04
Economics and roles
Trainer tiers by client count, a trainer's invited clients pay nothing, a referral programme with a discount and commissions. Every AI call is accounted for - unit economics without surprises.
/ 05
12 languages end to end
Not just the UI: the entire exercise catalogue and system messages are translated. A trainer in Tbilisi and a client in Berlin see the product as native.
- Outcome
What changes in the trainer's work.
x5
Clients per trainer. Was 5-8 in parallel - now up to 40 without losing quality. Workouts, nutrition and sessions - in one system.
12 languages
Localisation from day 0. The UI, the exercise catalogue and system messages - in full.
379 commits
In seven weeks - from a skeleton to 51 screens, 16 domains and 10 background services. The product grows in waves, each one live.
How it looks in the app.
The Telegram app: training, nutrition, diet protocols with an auto shopping list, and the exercise library - all in one dark interface.
Trainer, fitness studio, or expert running 1-on-1 practice - let's talk.
If you have a fitness studio, an independent practice, or an online school - I have a ready product, adaptable to your brand. And I have experience building the same kind of platform turnkey for adjacent niches: nutrition, therapy, coaching, healthcare.
Message on WhatsApp