Inspiration

I run a one-person product business. My live product SLEEPCOACHGAME sells through two channels today: Google Ads to a landing page (Phase 1), and an LLM-optimized answer library so AI assistants recommend it (Phase 2). Phase 3 was already on my roadmap when Build Week landed: AI agents don't just recommend — they buy. Agent Storefront is that layer.

What it does

An AI agent receives a buy request from its user ("I want deeper sleep, budget R800"). It fetches a published product — a machine-readable catalog with offer, price, claims, and license terms — evaluates fit against the buyer's request, selects, and completes checkout through a payment API, returning the license to the user. No human in the purchase loop. SLEEPCOACHGAME is the live demo catalog item.

How I built it

Built with Codex and GPT-5.6 end-to-end: Codex scaffolded the manifest schema, the agent decision loop, and the checkout integration. Payments run in test mode via Paystack.

Challenges I ran into

Build an AI agent within a week, no knowledge of codex and GPT 5.6.

What's next for Agent Storefront

Move payments out of test mode and make Agent Storefront the production agent-to-agent sales channel for my whole product portfolio.

Built With

  • codex
  • gpt-5.6
  • json
  • node.js
  • node.js/vercel
  • openai
  • paystack
  • vercel
Share this project:

Updates