DEV Community

DevEngine Labs
DevEngine Labs

Posted on

The Scaffolding Engine is Live — 128 Stacks, One Configurator, Zero Opinions

This is the follow-up to my earlier post about building an infinite full-stack assembly line. The engine is no longer "coming soon." It shipped.

Last week I wrote about a problem every developer knows intimately: you buy a boilerplate, spend the first three days tearing out the choices you didn't make, spend three more days figuring out why the pieces you kept are broken, and by the time you're writing actual business logic you've lost a week of momentum.

I said I was building something different. Here's what actually shipped.

What DevEngine Labs Does

You pick one option from four layers:

Frontend: Next.js, React SPA, Nuxt, or SvelteKit
Backend: Express, FastAPI, Fastify, or NestJS
Auth: Clerk or Auth.js
Billing + Database: Stripe + Supabase, Stripe + Firebase, LemonSqueezy + Supabase, or LemonSqueezy + Firebase

The engine pulls your four selected modules, compiles them into a single zip, generates a README written specifically for your combination, and delivers it. The auth is already wired to the backend. The payment webhooks are already connected. The database client is already initialized.

You unzip, drop in your API keys, run install, and you're writing features.

That's 4 × 4 × 2 × 4 = 128 distinct, pre-tested combinations — all available right now.

Why This Architecture Works

The key insight is treating each layer as a strictly decoupled module with enforced boundary contracts. The FastAPI backend module doesn't know or care whether it's being paired with Next.js or SvelteKit. The Clerk auth module doesn't know whether it's sitting on top of Supabase or Firebase. Each module exposes a clean interface and the compiler handles the wiring.

This is why adding a single new module multiplies the library rather than adding to it. When Prisma and Drizzle drop in as database/ORM modules in the next phase, the combination count doesn't go from 128 to 130 — it goes to 512+. The architecture scales geometrically.

The Stack Nobody Else Supports

Every major boilerplate on the market is a Next.js product. Some of them are excellent Next.js products. But if your MVP calls for a Python backend — and if you're building anything AI-adjacent in 2026, it probably does — you're on your own.

FastAPI + any frontend is completely unserved by the current boilerplate market. Same with Nuxt. Same with SvelteKit. Those communities are starved for production-ready starting points.

DevEngine is the only place I'm aware of where you can get a pre-wired Nuxt + FastAPI + Clerk + Stripe stack as a downloadable zip. If that combination describes your next project, it exists. Right now. Ready to run.

What's Coming

The next phase is the database expansion loop I wrote about in my first post. Prisma and Drizzle are already in testing as injectable ORM modules. When they land, the library hits 512+ combinations.

After that: multi-backend configurations. Some architectures genuinely need a Node.js API layer handling real-time operations alongside a Python service handling ML inference. That configuration will be selectable as a single purchase.

The economics of automated module compilation make it possible to keep adding layers without the maintenance overhead that would make this unworkable as a manually-maintained template library.

Try It

The configurator is live at dev-engine-labs.vercel.app.

Pick your stack. If your combination exists — and there's a 100% chance it does (or will very soon) — you'll have it in your inbox before you finish reading this sentence.

One-time purchase. No subscription. No opinionated leftovers.

Top comments (0)