Marea Alcalina

LATAM merchants don't want to wrestle with a CMS. They needed to stand up a digital menu and a storefront in minutes, not weeks, and edit them from a phone in the middle of a busy shift. The challenge was hiding all the complexity of serious e-commerce behind something as simple as sending a message.

At the same time, the platform had to perform like a real store: pages that load instantly for SEO and conversion, reliable cart and checkout, and catalogs that grow to thousands of products without degrading. Serving that to 200,000+ users demanded server-side rendering and infrastructure that scales on its own.

And the AI couldn't be an afterthought bolted on at the end. It had to actually edit —content, design and products— with human-level precision, and be robust enough to operate without constant supervision.

We built the front end in Angular 18 with server-side rendering on Cloud Run: every store is served pre-rendered so it loads fast and ranks, then hydrates for the editor's interactivity. The result is a storefront that feels instant and is fully editable at the same time.

Marea AI runs on OpenAI and Claude through 23 tools exposed by an MCP server. Each tool is a real platform operation —create a section, change a block, publish a product— that the model invokes while chatting over web or WhatsApp. That same MCP layer opens the door to external agentic clients.

Firebase handles data, auth and files; Stripe Connect runs per-merchant billing with payment splits. And 3,000+ automated tests harden every critical flow —checkout, editing and AI— so we can move fast without breaking production.