Most weekend MVP guides cheat. They count from a working boilerplate and skip the part where Friday night gets eaten by auth setup. So here's the honest hour-by-hour walkthrough of a real weekend build that ends with someone able to enter a card and pay you.
The stack is Next.js 15, Supabase, Stripe, and Cursor. No new frameworks, no obscure picks. The boring choice ships.
The weekend isn't the constraint. The scope is. The constraint is what you let yourself ship without.
The Weekend MVP Constraint And Why Most People Miss The Scope
A weekend is roughly twelve usable hours. That's two on Friday night, six on Saturday, and four on Sunday before fatigue wins. Anyone who pretends it's more is lying about caffeine.
Twelve hours buys you one job-to-be-done with a paywall and a landing page. Not a dashboard with five views, not a settings page with seventeen toggles, and absolutely not a multi-tenant team feature. One workflow that does one thing for one user.
The mistake that kills weekend MVPs isn't the tech. It's scope greed. You look at your favorite SaaS and think "surely I can copy that on a Saturday." You can't. They have a team, six months, and a designer.
Friday Night Two Hours: Repo, Auth, Database
Friday is plumbing, not product. The goal here is to wake up Saturday with a working login, a deployed homepage, and a database that already has your tables.
Spin up a Next.js 15 app with the TypeScript and Tailwind flags. Push to GitHub. Connect it to Vercel inside the same minute, because the first Vercel deploy is the most motivating commit you'll ever make.
Then go to Supabase and create the project. Copy the URL and anon key into your .env.local. Run the official Supabase Auth helpers install. By 9:30pm you should have a Google sign-in button that works locally and a row appearing in auth.users when you click it.
The Next.js Plus Supabase Plus Stripe Foundation
This trio survived the past four years because each piece does one job exceptionally well. Next.js handles routing and rendering, Supabase handles data and auth, Stripe handles money. Anything more, and you're juggling tabs you don't need.
Don't waste time picking ORMs. The Supabase JS client returns typed rows when you point it at your schema, and that's enough for a weekend. Drizzle and Prisma are great but they cost you forty-five minutes of setup you don't have.
For UI, install shadcn/ui and pull in maybe six components. Button, Card, Input, Dialog, Sonner for toasts, and DropdownMenu. That's the whole library you need this weekend.
The temptation to pick a fancier auth library, a stricter ORM, or a hipper component kit is the single most expensive trap of weekend MVP building.
Cursor And Claude Code As The Force Multipliers
The honest reason a weekend MVP is possible in 2026 and wasn't in 2022 is that AI tools write boilerplate faster than you can think it. Cursor for the IDE work, Claude Code for the larger refactors and the docs.
Cursor's agent mode handles one-shot file edits cleanly. "Add a server action that creates a project row scoped to the current user" is the kind of prompt that produces a clean, paste-ready file in under thirty seconds.
Claude Code earns its keep on Sunday when you start touching three files at once. The terminal agent reads your repo, makes a plan, and edits cleanly without breaking imports. We tested it on a four-feature ticket and the diff quality was sharper than any IDE plugin we tried, which lines up with what we found in the best AI coding assistants roundup.
Saturday Six Hours: Core Workflow End To End
Saturday morning starts with the smallest version of your core workflow that still does something useful. If your SaaS is an AI summarizer, the workflow is: user pastes text, clicks button, sees summary. That's it. No history, no folders, no sharing.
Get the input form working in the first hour. Then wire the API route that calls whatever third-party service does the actual work, whether that's OpenAI, Claude, Replicate, or your own logic. Persist the result to Supabase so the user can see it again on refresh.
Hours three and four go to the read view. A simple table or list that shows what the user has created. By 4pm Saturday you should have a working flow from input to output to retrieval, even if it looks rough.
The last two Saturday hours are styling and dead-link cleanup. Nothing fancy. Make the buttons look like buttons, the headings hierarchical, and the empty state friendly. Use the Tailwind defaults and you'll be fine.
Saturday Night One Hour: Deploy To Vercel
The Vercel deploy at 9pm Saturday is where most weekend builds break, because environment variables get forgotten. Open the Vercel dashboard, set every env you have in .env.local to Production and Preview, then redeploy.
Hit your production URL. Sign in with the Google account. Run the workflow. If it works in production, you've earned Sunday. If it doesn't, the bug is almost always one of three things: a missing env, a hardcoded localhost URL, or a Supabase Row Level Security policy that blocks the production session.
If you ship nothing else this weekend, ship the Saturday-night deploy. A URL someone can hit on Monday is what separates this build from the seven you didn't finish.
Sunday Three Hours: Stripe Checkout And Paywall
Sunday morning is Stripe. Create the product, copy the price ID, install @stripe/stripe-js and the Node Stripe SDK. Set up the Checkout session endpoint. Wire the success URL back to your app with a session ID query param.
The webhook is the part that always takes longer than expected. Use the Stripe CLI to forward events to your local /api/stripe/webhook while you build it. On checkout.session.completed, update a subscriptions table in Supabase with the user's plan and active flag.
The paywall itself is one server-side check. If user.subscription.active is false, redirect to /pricing. If true, let them through. Two lines of middleware code, not a whole library.
Don't bother with Customer Portal on weekend one. Send canceled users a manual email. You can wire the portal in week two. Same with annual plans, coupons, and tax. Default to monthly and one tier.
Stripe alone solves the global tax thing if you're an indie hacker shipping to one country. The moment you go international, you'll regret not picking Lemon Squeezy or Paddle from day one.
Sunday Hour: Landing Page Plus Three Validation Emails
By Sunday afternoon the app works and takes money. What's left is the landing page and the validation push. The landing is one screen: headline, sub, two screenshots, pricing block, signup button.
Write the headline as a problem statement, not a feature list. "Stop paying ten interns to write blog briefs" beats "AI-powered content brief generator." The first one says what the user gets, the second says what you built.
Then send three emails. One to a friend who's the target user. One to a Slack or Discord community where the target hangs out. One reply to a relevant tweet that already has engagement. Three messages on Sunday afternoon is enough to know if the idea has a pulse.
What We Ripped Out To Ship On Time
The honest list of what we cut from every weekend MVP we've ever shipped: profile pages, password reset emails (Google OAuth handles it), team invites, billing history, dark mode, custom domains, and any setting that isn't critical to the core workflow.
We also cut analytics on launch weekend. Posthog and Plausible can wait until Monday. You're not learning anything from analytics on day three anyway. You're learning from the three people who replied to your emails.
One more thing we always cut. The custom design system. shadcn/ui defaults are fine for v1. The day you have ten paying customers asking for a feature, you can hire a designer or rebuild the look. Until then, every hour spent on a custom Tailwind theme is an hour not spent on the workflow that actually takes money.
The Three Things That Always Break On Weekend Builds
Pattern one. The Stripe webhook works locally but fails in production because you forgot to add the webhook endpoint in the Stripe dashboard for live mode. Fifteen minutes of debugging if you know what to look for, two hours if you don't.
Pattern two. Supabase Row Level Security blocks your queries in production. The default policies are restrictive, which is the right default but creates the most common weekend bug. Write a single permissive policy per table on weekend one and tighten on week two.
Pattern three. Vercel function timeouts on the AI API call. The default is 10 seconds on the free tier, which is fine for most use cases but fails on long OpenAI or Claude calls. Upgrade to Pro before launch weekend if you're calling AI APIs directly.
Each of these is preventable if you know to look. Each of them eats your Saturday evening if you don't. The lesson from every weekend build is that the bug is rarely in your code; it's in the boundary between your code and a third-party service.
For the deeper Stripe versus alternative story when you start hitting global tax, look at the planned Stripe versus Paddle versus Lemon Squeezy comparison. And if you'd rather pull from a managed Postgres alternative, the Supabase versus Firebase breakdown covers the same trade-off.
Frequently Asked Questions
Can a non-developer really ship a SaaS in a weekend with this stack?
Not quite. The hour-by-hour assumes you can read JavaScript, debug a 500, and read a Supabase RLS policy. If you've never seen a terminal, you'll spend the weekend setting up and not shipping. Try Lovable or Bolt instead and read the vibe coding comparison.
Why Next.js over Remix or SvelteKit?
Vercel deployment is friction-free, the docs are exhaustive, and most AI coding tools default to it. SvelteKit is faster to write but harder to get help on at 2am.
Is the Supabase free tier really enough?
For weekend launch, yes. The 500MB database and 5GB bandwidth ceiling holds for hundreds of test users. The day you have real traffic, the $25 Pro tier solves it. See the official Supabase pricing.
What about authentication providers besides Google?
Add them later. Google covers eighty percent of indie SaaS signups and Supabase Auth lets you add GitHub, Apple, or email in five minutes when you need it.
Should I use a boilerplate like ShipFast?
If you've never built this stack before, yes, the $199 saves you Friday night. If you've built one before, you'll outpace the boilerplate by Saturday morning because you know exactly which parts to skip.
How do I handle the EU VAT and US sales tax problem?
Stripe Tax is the cheapest path if you stay on Stripe. Otherwise, switch to a merchant of record like Paddle or Lemon Squeezy. Plan for this on week two, not weekend one.
What happens to the MVP after the weekend?
Monday is for fixing the first bugs your three validators reported. Week two is for the one feature they asked for that you can build in a day. If nobody replied, the idea didn't have a pulse and your weekend cost you nothing.
Is Claude Code really better than Cursor for this kind of work?
For one-file edits, Cursor wins. For three-file changes and refactors, Claude Code wins. Most weekend MVPs lean Cursor because the work is small-scope and visible in one editor pane. The Anthropic Claude Code docs walk through both modes.
Can I skip Vercel and self-host?
You can, but not on launch weekend. Coolify on Hetzner is brilliant for week three. For weekend one, Vercel's zero-config deploy buys you back two hours you don't have.
