TarotStory

AI tarot readings in a voice the founder controls

TarotStory is a subscription web app where a language model writes a personal tarot reading for the question you ask, once a day. The interesting part is who controls it: the founder writes and edits every instruction the model gets, from its persona to which questions it turns down, in a backoffice and without a developer. We built the product from the client's Figma designs: the web app, the backend, payments and memberships, narration, and that backoffice.

AI tarot readings in a voice the founder controls
What we did
  • Web App Development
  • Backend Architecture
  • AI Integration
  • Payments & Memberships
  • Backoffice
  • QA Engineering
Built with
  • Next.js
  • TypeScript
  • Supabase
  • Stripe
  • Vercel

The Challenge

An AI that answers personal questions will be asked things it should not answer: a diagnosis, a court case, an investment, or a question from someone in crisis. The founder also needed to own the reader's voice and keep tuning it after launch, without a developer in the loop for every change. And a daily reading only becomes a habit if the product remembers what came before.

Our Approach — from the TarotStory build

Our Approach

Every question is screened before a card is dealt: medical, legal, financial, crisis and self-harm questions get a gentle pause and an invitation to rephrase, and if the check cannot run, nothing is sent. The persona, the instructions for each stage of the reading and the screening rules live in the backoffice and are versioned on every save, so an edit applies from the next reading. Numerology readings send the model the numbers only, never the user's name or date of birth. Saved readings collect into Stories, where each week of readings becomes a chapter, written overnight, that picks up where the last one left off.

The Result — from the TarotStory build

The Result

TarotStory is built, tested end to end and deployed, and is in its launch phase with the waitlist open at www.tarotstory.com. The founder tunes the voice and the screening rules directly in the backoffice, and each change is live from the next question.

Want a build like this one?

Same team, same senior engineers. Tell us what you are building and we will tell you how we would approach it.

Google Cloud Partner

Tell Us About Your Project