Ship a UK SaaS in a weekend with Claude Code + Vercel + Supabase: the Friday-to-Sunday playbook with GBP costs and the four mistakes that kill the timebox

Two days. One weekend. A real UK SaaS - paid, deployed, smoke-tested, indexable - by Sunday evening. The Claude Code + Vercel + Supabase stack is the de facto UK indie hacker default in 2026 because it does the boring stuff (auth, DB, hosting, CI/CD) for under GBP 30/month and gets out of your way. This is the Friday-evening-to-Sunday-night playbook: what to ship in each phase, what to skip, and the four mistakes that turn a weekend build into a fortnight.

What you'll have by Sunday night

A live URL on a custom UK domain. Email/password (and magic-link) auth. A Stripe checkout in test mode that successfully takes a GBP 9.99 payment and writes a row to your database. A core feature loop a real user can actually use. A simple landing page with a sign-up form. Indexed by Google within 48 hours. Total monthly cost: under GBP 30 including the Anthropic subscription.

Realistic? Yes, for a single-feature SaaS with one paying user persona. You will not ship a full multi-tenant CRM by Sunday. You will ship the version of one that has one feature, real auth, real billing, and a real first user.

The stack and why it works in two days

Layer	Tool	Tier	GBP/month
Code agent	Claude Code	Pro	16
Hosting	Vercel	Hobby	0
Database + auth	Supabase	Free	0
Payments	Stripe	Pay-as-you-use	1.4% + 20p per UK card
Email	Resend	Free (up to 3k/month)	0
Domain	123-Reg or Cloudflare	.co.uk	~ 8/year
Total monthly fixed			~ 16

The combination works because each tool ships with batteries included for the others. Supabase has a Vercel deploy button. Stripe has a Supabase template. Claude Code's context is big enough to hold the whole stack. There's no glue code because the glue is documented and Claude Code already knows it.

Friday evening (3 hours): scaffold and deploy

The first session is about getting a deployed shell of an app on a real URL. Don't write features. Don't customise styling. Just get from empty folder to live URL with a working Hello world.

6:00pm — Domain. Register your .co.uk on 123-Reg or Cloudflare Registrar. About GBP 8/year. Don't overthink it; you can rebrand later. While DNS propagates, do the rest.

6:15pm — Project init. Make a folder, run claude inside it, and prompt:

"Scaffold a Next.js 15 app with TypeScript, Tailwind, the Supabase JS client, and shadcn/ui. Set up a basic /, /login, /signup, and /dashboard route. Add a CLAUDE.md documenting the stack. EN-UK spelling throughout."

Five minutes later you have a running Next.js app.

6:30pm — Supabase project. supabase.com → New project → free tier → London region. Copy the project URL and anon key into .env.local. Run the auth-helpers setup. Tell Claude Code: "Wire Supabase auth so /dashboard requires a session and redirects unauthenticated users to /login."

7:00pm — First commit + Vercel. GitHub repo, push, then vercel link and vercel --prod. Set the environment variables in the Vercel dashboard. Within ten minutes you have your-app.vercel.app live.

7:30pm — Custom domain. Add your .co.uk in Vercel. Update your DNS records. Wait for the certificate to provision (usually under 5 minutes).

8:00pm — Sanity check. Sign up on the live URL with a real email. Check that a row lands in auth.users in Supabase. Check that the /dashboard route is gated. If both work, you're done for Friday. Have a beer.

You should not have written a feature yet. If you did, you've over-shipped Friday. The point of Friday is to remove every infrastructure question before Saturday morning.

Saturday (8 hours): the core feature loop and Stripe

Saturday is the day. You're going from working-shell to working-product. Two big tasks: ship the feature, wire up Stripe.

9:00am — Feature spec. Before you open Claude Code, write down what your one feature does in three sentences. "User uploads a UK postcode. App returns a 12-month rainfall forecast. The forecast is saved to their account history." That's it. If you can't write the feature in three sentences, your weekend scope is too big.

9:30am — Schema and core endpoint. Tell Claude Code:

"Add a Supabase table called forecasts with columns id (uuid), user_id (uuid foreign key to auth.users), postcode (text), forecast_json (jsonb), created_at (timestamptz). Add a Row Level Security policy so users can only see their own rows. Then add an authenticated API route at /api/forecast/[postcode] that calls the Met Office DataPoint API, stores the result, and returns it."

This is one round-trip if your prompt is specific. Two if you forgot to mention RLS the first time.

11:30am — Frontend. A single page that takes a postcode, calls the endpoint, shows the result. Tell Claude Code: "Add a /dashboard page with a postcode form, a loading state, and a results card. List the user's previous forecasts below."

1:00pm — Lunch break. Don't skip this. Tired prompts lead to bad code.

2:00pm — Stripe setup. Stripe dashboard → create a product (Pro plan, GBP 9.99/month). Get the price ID. Tell Claude Code:

"Wire up Stripe checkout with the price ID price_xxx. Add a /pricing page with a CTA that creates a checkout session. Add a Stripe webhook receiver at /api/webhooks/stripe that listens for checkout.session.completed and updates a subscriptions table in Supabase with the user's plan and status. Verify the webhook signature."

4:00pm — Test the full flow. Use Stripe's test card 4242 4242 4242 4242. Sign up, check out, get redirected back, see your subscriptions row. If this works, you're 80% done.

5:00pm — Gate the feature. Add a Pro check on the postcode endpoint. Free users get one forecast a day; Pro users get unlimited. Tell Claude Code; it'll wire it in 5 minutes.

6:00pm — Marketing landing page. A single landing page on / with the hero (problem, promise, CTA), three benefits, pricing, and a sign-up button. Tell Claude Code: "Build a clean, conversion-focused landing page using shadcn/ui, EN-UK copy, GBP pricing." Don't waste the evening polishing it - good enough on Saturday is great by Wednesday.

8:00pm — Ship Saturday's work. Push to main, watch Vercel deploy. Sign up on production with a fresh email, check out with the test card (in Stripe test mode), confirm everything works end-to-end on the production URL.

Sunday (4 hours): production switchover and smoke test

Sunday morning is the production cutover. Sunday afternoon is the smoke test and first marketing post.

10:00am — Stripe to live mode. Switch your Stripe keys from test to live. Update Vercel env vars. Redeploy. Re-test with a real card (use your own; refund yourself afterwards).

11:00am — UK VAT decision. If you're not VAT-registered, charge GBP 9.99 inclusive and don't collect VAT. If you are (or will be), Stripe Tax handles the calculation - turn it on, set your origin to the UK, and Stripe figures out place-of-supply. We've covered the VAT place of supply rules at length; for weekend zero, just take the money inclusive.

11:30am — Resend for email. Sign up at resend.com (free up to 3,000 emails/month), verify your domain (DNS records take ~10 minutes), and tell Claude Code: "Wire up Resend for transactional emails - welcome email on sign-up, payment-success email after Stripe webhook." Templates from react-email look fine out of the box.

12:30pm — Lunch.

1:30pm — Smoke test the production environment. Real card, real email, real flow. New incognito window: sign up, log in, run the core feature, check out for Pro, run the gated feature, log out, log back in, see your data. Anything that breaks - fix it now.

3:00pm — SEO basics. Tell Claude Code: "Add a sitemap.xml route, a robots.txt, JSON-LD organization schema, and Open Graph tags to every page. Title and description for the homepage and pricing page in EN-UK." Submit the sitemap to Google Search Console. Indexing usually starts within 48 hours.

4:00pm — First marketing post. A single tweet, a single LinkedIn post, a single Reddit comment in a relevant subreddit. Don't beg. Lead with the problem you solved, link the URL.

5:00pm — You're done.

What it costs

Item	First month	Notes
Anthropic Pro	GBP 16	Cancel anytime
Domain	GBP 8/year	One-off
Vercel	GBP 0	Hobby tier handles weekend traffic
Supabase	GBP 0	Free tier covers 500MB + 50k auth users
Stripe fees	1.4% + 20p per UK card	Only on actual revenue
Resend	GBP 0	Up to 3k emails/month
Total fixed	GBP 16	Plus GBP 8 one-off domain

You'll outgrow the free tiers when you hit ~50k Vercel page views or ~500MB of Supabase data. Both upgrade paths cost GBP 16/month each. So GBP 48/month total is the realistic upper bound for a SaaS doing low-five-figure MRR.

The four mistakes that kill the timebox

One. Scope creep before Friday's beer. "Should I add team accounts?" is a Sunday-week-three question. Friday says: one user, one feature, one price. If your first prompt to Claude Code includes the word "multi-tenant," your weekend is over.

Two. Rebuilding what Supabase already gives you. Supabase auth handles email/password, magic links, OAuth, password resets, email verification, RLS, JWT issuance. Don't write any of this. If you find yourself in auth/[...nextauth] files on Saturday, you've taken a wrong turn - delete and use Supabase's built-in helpers.

Three. Stripe in production before you've tested in test. Live keys before a clean end-to-end test in test mode is how you find out at 11pm Saturday that webhooks don't reach localhost. Test in test mode first, then switch.

Four. Letting Claude Code write 800 lines you didn't read. This is the agentic-loop trap. Big diff, looks impressive, contains three subtle bugs. Read every PR-shaped diff. Use /agents for read-mostly work; don't use it for write-mostly work without reviewing.

What to do on Monday

You shipped. Now: sleep, then on Monday morning answer one question - did anyone you don't know sign up over the weekend? If yes, you have a lead and a thread to pull. If no, your distribution is the bottleneck, not your build. Either way, the product exists, the meter is running, and your weekend cost was GBP 16 + GBP 8 + maybe GBP 12 of test transactions you'll refund yourself.

This is the iteration that compounds. Most UK indie hackers who shipped one weekend SaaS shipped a second within a month, because the muscle memory works.

Closing

The point of the weekend timebox isn't speed for its own sake - it's forcing scope honesty. Two days makes you cut every "nice to have" until you're staring at the actual core. The Claude Code + Vercel + Supabase stack works for this because nothing in the path requires you to make decisions you don't want to make on a Saturday afternoon.

If this is your first time using Claude Code, our first-week starter playbook is the right pre-read. If you're still picking your tool, the Cursor vs Claude Code UK buyer's guide gives the GBP-priced answer. And if you want VAT done properly before your first paying user, our VAT place of supply guide is twenty minutes well spent.

Free this week's data-backed UK builder report. Every Thursday IdeaStack publishes one deeply researched UK opportunity - keyword volumes, SERP gaps, GBP pricing, a builder prompt. Read the latest free report.