Nitro — AI Product Discovery for E-commerce

Online shoppers don’t always know what they want — Nitro turns that into a guided quiz that recommends the right product and quietly lifts conversion. The platform has to drop into any storefront and talk to every marketing tool a merchant already uses, which is a deceptively large engineering problem.

The challenge

“Universal” is the hard word. A recommendation widget that works on Shopify is one thing; one that behaves identically across Shopify, WooCommerce, BigCommerce, Magento, Wix, Squarespace, and a dozen more — each with its own theming, lifecycle, and quirks — is another. Most quiz tools cheat with an iframe; Nitro embeds natively for the sake of performance and conversion. On top of that, merchants expect their quiz data to flow into the 65+ marketing tools they run — Klaviyo, Attentive, GA4, Meta, TikTok, Recharge, Yotpo — without a developer in the loop. And the merchant dashboard has to make all of it legible.

The approach

We structured the whole system as a Turborepo monorepo of 26 packages so the platform SDKs, the dashboard, and the API evolve together and never drift. The API runs on Hono with TypeScript, backed by Prisma, PostgreSQL, and Redis for the recommendation hot path. A single normalized event pipeline fans out to the 65+ marketing integrations, so adding a destination is a connector, not a project.

What we built

• A zero-iframe universal embed (nitro-embed) plus a nitro-react component layer, so the quiz renders natively in any theme instead of inside a sandboxed frame
• 14 platform SDKs — nitro-shopify, nitro-woocommerce, nitro-bigcommerce, nitro-magento, nitro-wix, nitro-squarespace, nitro-saleor, nitro-medusa, nitro-commercetools and more — backed by real storefront apps (Shopify theme extension, WordPress plugin, Magento module, Wix Blocks app)
• A drag-and-drop quiz builder with conditional logic, and a probabilistic recommendation engine powered by nitro-vector embeddings
• A 65-integration layer (nitro-integrations) spanning email, SMS, analytics, ads, reviews, loyalty, subscriptions, and automation from one event schema
• The nitro-dashboard (Next.js) for quiz building, analytics, and integration management, plus a separate nitro-admin panel and nitro-docs site
• CAOS v2 AI agents (nitro-agents) for internal support, CS, and revenue ops, and a nitro-experiments package for built-in A/B testing

The outcome

Nitro proves a SaaS platform can be both genuinely universal and pleasant to operate. That’s the same discipline we bring to e-commerce builds of any size — serious infrastructure under an interface that just works.