Building VitrOS: From Zero to Production SaaS in 30 Days

Thirty days. Zero to production SaaS. Not a landing page. Not a Figma mockup. A real multi-tenant app with auth, payments, role-based access, offline PWA capability, and a database you can stress-test. I am going to walk you through exactly how I built VitrOS and what I would change if I started over.

Before the code: VitrOS is a visual ops system for small teams. Think Linear meets Notion but cheaper and focused on shipping. The market is saturated, which is fine, because my goal was not to win the market. My goal was to prove I could go from an empty repo to a paying customer in thirty days.

The Stack Decision

People love to debate this forever. I did not have time. Here is the stack I shipped with, and why.

• Next.js 16 with the App Router. React Server Components give you the best of both worlds: fast page loads and dynamic data without hydration nightmares.
• PostgreSQL hosted on Neon. Branching databases means I can spin up a staging copy in seconds.
• Prisma as the ORM. Yes there are faster options. No I do not care. Prisma has the best developer experience and the migration story is solved.
• NextAuth v5 for auth. Handles email, OAuth, and sessions without me touching a cookie.
• Stripe for payments. Stripe Checkout for the first paying customer, then migrated to a custom flow in week three.
• Tailwind 4 for styling. Design tokens as CSS variables. No CSS-in-JS.
• Vercel for deployment. Git push and it is live in 90 seconds.

The whole stack is boring and battle-tested. That is the point. When you are racing a deadline, boring infrastructure lets you spend your energy on the parts that actually matter: the product.

Week 1 — Schema and Auth

The temptation on day one is to start wireframing. Do not. Start with the database schema. Everything else downstream depends on it. If you get the schema wrong, you will refactor for weeks.

VitrOS has five core tables: User, Workspace, Membership, Project, and Task. The Membership table is the key to multi-tenancy. Every query in the app joins through Membership to filter by workspace. If you skip this step you will build a single-tenant app by accident and regret it later.

I spent the first three days on Prisma schema design. I threw away two drafts. Day four I was writing seed scripts and day five I had NextAuth v5 wired to Google and email. By the end of week one I had signup, login, workspace creation, invite flows, and role-based access control. Zero UI. Just database and middleware.

The multi-tenant pattern that saved me

Every query in the app runs through a helper called getScopedPrisma(userId, workspaceId) that checks Membership before returning a Prisma client. It is not the fastest approach but it is impossible to accidentally leak data across workspaces. I slept better because of it.

Week 2 — The Product UI

With the backend locked, week two was pure UI. I built the shell first: sidebar, top nav, workspace switcher. Then the core workflow: create project, add tasks, drag to reorder, mark complete. I used Framer Motion for the animations and shadcn/ui for the base components.

Here is the rule I follow in UI work: if it takes more than two hours to build a component, stop and find a library. Your job is to ship. It is not to reimplement a drag-and-drop library for the fifth time in your career.

Week 3 — Payments and Edge Cases

Payments are where most indie SaaS projects die. Not because Stripe is hard — Stripe is actually great — but because you have to think about so many edge cases. Upgrades. Downgrades. Cancellations. Failed charges. Prorations. Refunds. Trial expirations.

My shortcut: launch with monthly billing only, no trials, no coupons, no proration. One price. One plan. It is not optimal for revenue. It is optimal for shipping in 30 days. You can always add complexity later.

The Stripe webhook handler is the most important piece of code in a SaaS app and the one people always skip. Your database state must follow Stripe, not the other way around. When a webhook comes in, you update your local subscription state. When a user clicks cancel, you call Stripe and wait for the webhook to confirm. Do not trust the UI response.

Week 4 — PWA, Polish, and Shipping

Week four was the scary week. This is where most projects pile on last-minute features and end up with a broken launch. I forced myself to do the opposite: feature freeze on day 22, then only polish and PWA work.

The PWA piece was the most fun. Next.js 16 has first-class PWA support with a service worker that handles offline caching for the core routes. I cached the app shell, the user profile, and the most recent projects. If the user loses connection, they can still see their workspace and queue writes that sync when they reconnect.

The queue is a simple IndexedDB store. Every mutation goes into the queue first, then fires off to the server. If the server fails, the mutation stays in the queue and retries. If it succeeds, the queue entry is cleared. This gives VitrOS the feel of a native app without any of the native app overhead.

What I Would Do Differently

A few things I got wrong.

1. I used Prisma migrations in production too early. I should have used raw SQL for the first 30 days. Prisma migrations are great until you need to do a zero-downtime schema change at 2 AM. Use raw SQL until you are confident the schema is stable.
2. I skipped observability. No logging. No error tracking. When the first bug hit production I had to reproduce it locally from memory. Add Sentry or OpenTelemetry on day one, not day thirty.
3. I built the admin panel last. I should have built it first. Every SaaS needs a way for the founder to manually fix data, grant access, and debug users. Having to SSH into production to run queries on day 15 was not fun.
4. I underpriced the first plan. I launched at $12 per seat because I was scared to charge more. The first ten customers would have paid $40 per seat without blinking. Price reflects perceived value. Do not underprice out of fear.

The Receipts

30 days from empty repo to production. First paying customer on day 18. First $1k MRR on day 27. As of this post: $4,200 MRR and climbing. The full repo is available to Pro members along with the Prisma schema, the Stripe webhook handlers, and the NextAuth v5 config.

The real lesson is not the tech stack. The real lesson is that 30 days is plenty of time to ship something real if you stop treating infrastructure decisions as identity-defining. Pick boring tools. Ship. Iterate.