Light evidence (grade D: anecdotal)A short LinkedIn post by the owner. The cost figures are his own, and the adherence change isn't quantified.

A 112-Member Gym Swapped a $3,600 SaaS for Its Own App With an AI Coach

Persistence Athletics in Seattle replaced its $3,600-a-year Wodify subscription with a gym app built in Claude Code that runs on about $300 a year in Anthropic API calls and adds photo meal logging and an after-hours AI coach.

Who this is about: Gym (Persistence Athletics, Seattle, 112 members)

Claude Code ClaudeExpoStripeWHOOPLocal & service Solo foundersAutopilot

Get the prompt

ProductAdded Oct 15, 2026Source: linkedin.com

What it takes

Cost: PaidPay-as-you-go Claude API usage: about $300 a year, by the owner's figures.
Setup time: A few weeksOur estimate from the source, for one business.
Skill needed: Developer skillsYou build or run your own code, even if AI writes it.

Explain it simply

A gym was paying $3,600 a year for software to run memberships. The owner used Claude Code, an AI that writes software, to build the gym's own app, which costs about $300 a year in AI fees. It also lets members snap a photo of a meal to log what they ate and chat with an AI coach that knows their training.

New to AI? See how AI works

What was done

The owner built a gym operating system with Claude Code to replace Wodify, with Stripe billing and WHOOP integration. Members can snap a meal photo, and a vision agent identifies it, estimates macros, asks them to confirm and logs it in three taps. Members can also message an AI coach that knows their week of training, their last WHOOP recovery and the intent of the next session.

How it went

Software costs went from $3,600 a year to about $300 a year in API calls. Nutrition logging adherence went up the week the photo feature shipped. A member asking at 9:47 pm why their snatch felt slow gets an answer with the context the owner would have given in person the next morning.

The adherence change isn't quantified, and the owner says "the savings aren't the point." His main lesson: once building gets cheap, a clear spec matters more.
The post links to a fuller write-up we haven't reviewed.
Health and fitness advice from an AI coach should stay within your scope of practice.

Prompt

PromptWritten by our editors from the source

Meal logging agent: "Identify the foods in this photo, estimate portion sizes
and macros (protein, carbs, fat, calories). Show the estimate and ask the member
to confirm or edit before logging."

Coach agent context: the member's last 7 days of training, latest WHOOP recovery
score, and the intent of their next scheduled session. "Answer the member's
question using this context. If it's about pain or injury, tell them to talk to
a coach in person."

Our version; the post doesn't include prompts.

Business fit

Who this helps, and how:

Independent contractors (barbers, bartenders, consultants): give clients an app that answers between-session questions using their own history.
Local & service
Medical and health professional business owners: add photo-based logging for clients' meals or exercises, with a clinician checking anything outside a wellness scope.
Local & service
Local service businesses: replace a membership or booking subscription with your own app that includes Stripe billing.
Local & service
Content creators: offer paying members an AI coach that knows each member's progress.
Creators

Similar tools

Other tools for this job, and how they differ. Chosen by our editors, never paid for.

PushPress
Off-the-shelf gym management software, an alternative to Wodify without building your own.
Lovable
Builds apps from prompts with less hands-on coding.
Cursor
An AI code editor for building the same kind of app.

“Light evidence” (grade D) means: A brief, early or in-progress claim. Treat it as an idea to test, not proof. How we grade. Spotted a mistake? Suggest a correction.