Ship a UK micro-SaaS with Claude Code in a weekend: a UK-first playbook

The weekend-SaaS myth used to be exactly that. A myth. You'd read a breathless Twitter thread about some bloke in San Francisco who shipped a PDF summariser between his Friday oat flat white and his Sunday brunch, and you'd open your laptop full of hope. Three hours later you'd still be fighting with auth, the payment integration was fake, and the cookie banner was some nightmare you'd pasted in from 2019. Come Monday morning, you'd have a half-finished repo and a slightly bruised ego.

In 2026 that's genuinely changed. Claude Code can scaffold a real product in hours, not days. The stack has consolidated around a small number of AI-friendly tools that actually talk to each other. And if you're UK-based and you design for UK users from line one of your CLAUDE.md file, you skip the painful "I've built a great product, now let me rewrite half of it for British customers" step that ruined a lot of weekends in 2024 and 2025. This is the playbook for doing it properly — end to end, in a weekend, with real GBP customers by Sunday night.

What "UK micro-SaaS" actually means here

Let's define the thing before we build it. A UK micro-SaaS in this piece is a small, focused software product that:

• Charges between GBP 19 and GBP 49 per month
• Solves one specific problem for one specific user (B2C or B2B)
• Is billed in GBP from day one via Stripe GB
• Complies with UK law out of the box — no hand-wavy "we'll sort ICO later"
• Ships on a weekend using Claude Code plus one modern, AI-friendly stack

It's not Zendesk. It's not the next Notion. It's the kind of thing one person maintains from a spare bedroom in Leeds or Brighton and makes GBP 1,500 to GBP 8,000 a month in pure margin. The kind of thing IdeaStack's weekly reports tend to surface — small, specific, UK-native niches the US giants won't bother with.

The weekend schedule, hour by hour

Block it out on Friday morning. Tell your partner, your housemates, your cat, your fantasy football chairman. This is the shape.

Friday evening: 2 hours — pick the idea and write the CLAUDE.md

Don't pick the idea on Friday evening. That's a trap. Pick it earlier in the week from your backlog (or from this week's IdeaStack report) and walk into Friday night knowing exactly what you're building.

Friday's job is one thing: open a fresh folder, git init, and write the CLAUDE.md. This file is the single most important thing you'll do all weekend. It's Claude Code's brain for the next 48 hours. If you skip it, you'll end up with USD pricing, MM/DD dates, American spellings, and a cookie banner that thinks it's in Kansas.

Saturday morning: 4 hours — scaffold the stack

Coffee. Proper one, not instant. Then:

• Ask Claude Code to scaffold a Next.js 16 app with the App Router, TypeScript, Tailwind, and shadcn/ui
• Wire up Supabase for auth and the database
• Drop in Stripe GB with GBP as the presentment currency
• Add Resend for transactional email

Claude Code will do most of this in a single long session. You're the pilot. You read the diffs, you push back when it does something daft, you commit every time a feature works. Don't let it run wild — review every file before you accept it.

Saturday afternoon: 3 hours — core feature and onboarding

This is the feature that actually does the thing. One feature. Not five. If you can't describe it in one sentence, it's too big.

Build it, wire it to the database, add the onboarding flow (email, verify, land on the dashboard, see the first "wow" moment within 60 seconds). Onboarding is the bit most weekenders skip. Don't skip it.

Saturday evening: 2 hours — UK billing and cookie banner

Stripe GB, GBP prices in pence, UK VAT-aware receipts (even if you're under the threshold — your receipts should still make sense to a sole trader accountant), and a DUA-compliant cookie banner if you're using anything other than cookieless analytics.

If you stick to Plausible or Vercel Analytics for first-party cookieless tracking, you don't need a banner under the DUA Act 2025 exemption for first-party analytics. That's a genuine time-saver. Take it.

Sunday morning: 3 hours — polish and copy

The homepage, the pricing page, the legal pages. Claude Code can draft all of these but you need to actually read them. British English. GBP. Plain language. No "Optimise Your Synergies" corporate waffle.

Generate a hero image with whatever image tool you fancy (DALL-E, Ideogram, Midjourney). Keep it clean. Avoid the AI-generated-looking ones — they telegraph "made in 20 minutes" in a way that hurts trust.

Sunday afternoon: 3 hours — deploy and launch post

Vercel deploy. Wire the custom domain (register one from Namecheap or IONOS that morning if you haven't — give DNS time to propagate). Set your environment variables properly. Run the full flow yourself: land, sign up, pay, use the feature, cancel, re-subscribe. If any step breaks, fix it before you tell a single person about the product.

Write a launch post. Short, honest, tells the story of the weekend. You'll post it on LinkedIn, the UK indie-hackers channels, and anywhere your ideal customer spends time.

Sunday evening: 1 hour — ship to first 10 UK beta users

You should already have 10 names in a text file. People who said "yeah that sounds useful" when you mentioned the idea over the past week. Send each of them a personal message. Not a newsletter — a proper, typed-out note. Ask them to try it. Ask for feedback. Offer the first three months free in exchange for honest, detailed input.

By the time the football highlights are on, you've shipped.

The CLAUDE.md you should write on Friday

Here's a real one. Drop it in the root of your repo before you ask Claude Code to do anything.

# CLAUDE.md — UK Micro-SaaS Guardrails

## Locale
- All user-facing copy in British English (organise, colour, analyse, behaviour)
- Currency: GBP only, formatted as GBP 19 or £19 — never $ or USD
- Dates: DD/MM/YYYY in UI, ISO 8601 in the database
- Times: 24-hour clock where possible, London timezone as default

## Stack
- Next.js 16 (App Router, TypeScript, Turbopack)
- Supabase (auth + Postgres + RLS enabled from day one)
- Stripe GB account, GBP presentment, pence in the database
- Tailwind + shadcn/ui
- Plausible for analytics (cookieless, first-party)
- Resend for transactional email

## Compliance (non-negotiable)
- Must register with ICO before first paying customer (GBP 40 to 60/year)
- DUA Act 2025: cookieless analytics only by default — no cookie banner needed
- If any third-party tracker is added, a proper banner is mandatory
- Consumer Rights Act 2015: 14-day cooling-off on B2C digital subscriptions
- HMRC MTD: if Ltd and VAT-registered, invoices must be MTD-compatible
- Privacy policy and T&Cs written in plain English, linked in the footer

## Billing
- Prices in Stripe held as integer pence
- Default plan: GBP 19/mo, GBP 190/yr (2 months free)
- All invoices include a UK address, VAT line (even if zero-rated), and sole-trader-friendly numbering
- Webhooks are idempotent — check event.id against a processed-events table

## Branding
- No em dashes. Short sentences. Dry humour allowed.
- No US spellings anywhere in the product or marketing copy.
- GBP symbols, not dollar signs, even in example screenshots.

Thirty-odd lines. Saves you about six hours of arguing with Claude Code about why it keeps writing "color" and "$29".

UK compliance guardrails you must hit this weekend

Most weekend builds quietly skip the legal bit and hope for the best. In the UK that's a bad bet — the ICO has been visibly more active since 2024, and the new Data Use and Access Act 2025 clarified cookie rules in a way that actually helps small builders if you lean into it.

Here's the short list for a paying micro-SaaS:

ICO registration. If you process personal data (and you do — at minimum email addresses), you must register with the ICO as a data controller. GBP 40 to 60 a year depending on your size. It takes 20 minutes online. Do it before your first paying customer, not after.

DUA Act 2025 first-party analytics exemption. If you use cookieless first-party analytics like Plausible or Vercel Analytics, you are genuinely exempt from the cookie-banner requirement for that category. That means if you ship only with cookieless analytics and no marketing pixels, you don't need a cookie banner at all. This is a material simplification — take advantage of it.

Cookie banner, if you use Google Analytics, Meta Pixel, or similar. Claude Code can hand-roll you a compliant banner in about 30 minutes. Accept / reject / customise, granular categories, stores the preference in a first-party cookie, works on mobile. Don't use a cheap off-the-shelf banner that defaults to opt-in for everything — that is no longer compliant.

Privacy policy in plain English. Not lifted off a generator. Claude Code writes a perfectly good one from a 10-line spec: what data you collect, why, where it's stored, how long for, how to contact you, how to delete an account. Sign off on it yourself — it's your name on the tin.

Consumer Rights Act 2015 cooling-off. B2C digital services have a 14-day right of cancellation unless the customer explicitly waives it (and you keep evidence). For a monthly SaaS with a free trial this is usually fine — the trial period covers you. For annual plans, offer a clear 14-day refund window. Most customers never use it.

HMRC implications. If you're a sole trader and the side-hustle trading allowance is still GBP 1,000, you can earn up to that threshold before declaring. Above it, self-assessment. If you expect real traction, consider incorporating as a Ltd before revenue gets awkward. If VAT-registered, MTD for VAT is the reality — your invoices and Stripe exports need to be MTD-compatible. Talk to an accountant for two hours before you launch your second product. It's the best GBP 200 you'll spend all year.

UK Stripe GB integration, done properly

Stripe GB is the bit people fudge. Don't fudge it.

• Currency. Create all products and prices in GBP. Hold amounts in pence as integers. GBP 19.00 is 1900, not 19.
• Presentment. Set GBP as your default presentment currency in the Stripe dashboard. If you accept USD customers later, fine, but default to GBP.
• Dates. Your invoices and receipts should show DD/MM/YYYY. Stripe's default locale for a GB account handles this if you set customer preferred_locales to ["en-GB"].
• Radar rules. Turn on the standard UK fraud ruleset. Add custom rules for impossible travel and high-risk email domains if you're hit by card testing early.
• Webhook idempotency. Every webhook handler checks event.id against a processed_stripe_events table. You log, then process, then mark complete. This is 30 lines of code and it prevents the double-charge nightmare that eats weekends.
• Tax. Stripe Tax for GB is worth the fee if you're Ltd and VAT-registered. For sole-trader sub-threshold, skip it and add a simple "VAT not charged — seller below UK VAT threshold" line on your receipts.

Ask Claude Code to wire all of this on Saturday evening, and push it to test mode. Buy your own product with a test card. If the receipt looks like something a British accountant would accept, you're done.

The stack cost — real GBP numbers

Weekend builders get scared off by imagined cloud bills. Here's the honest breakdown for a from-zero weekend build:

• Vercel Hobby. Free. Covers everything a single-operator side project does in its first year.
• Supabase Free tier. Free. 500MB database, 1GB storage, 2GB bandwidth. More than enough for first 100 users.
• Stripe GB. No fixed monthly fees. You pay 1.5% plus 20p per UK card transaction. Zero if you have zero customers yet.
• Claude Code API usage. Roughly GBP 3 to 7 across the whole weekend for a normal build. Worst case I've seen personally was GBP 12 on a particularly chatty session.
• Resend Free tier. Free for the first 3,000 emails a month.
• Plausible. Free if you self-host, or GBP 7/month for hosted (worth it for a weekend build — self-hosting is a time trap).
• Domain. GBP 8 to 12 a year from a decent UK-friendly registrar.
• ICO registration. GBP 40 to 60 a year. Not optional.

Total out-of-pocket for the weekend itself: GBP 0 to 15 (Plausible subscription + domain if you pay annually). Plus the ICO fee before you take your first paid customer.

That's it. Bring back the scaremongering.

Five failure modes, and how to dodge them

1. Scope creep. You planned one feature. By Saturday afternoon you're "just adding" a second one. Stop. Write the second one down in your backlog. Ship the first one. You have seven hours between Sunday morning and Sunday evening and they are not for new features — they are for polish, deploy, and getting users.

2. Mocked-not-integrated Stripe webhooks. Claude Code will happily stub a webhook handler that logs "OK" and returns 200 without actually updating your database. You'll think billing works. It won't. On Saturday evening, buy your own product with a test card and verify the row actually appears in your subscriptions table. If it doesn't, fix it before bed.

3. No ICO registration. Easy to forget, easy to fix, but if you forget and your first paying customer asks "are you registered with the ICO?" you look amateur. Do it Friday night. It takes 20 minutes.

4. Cookie banner that breaks on mobile. If you added one, test it on an actual phone, in portrait, over a flaky 4G connection. Too many weekend banners cover the CTA button on mobile and tank conversion. Better still: stick to cookieless analytics and avoid the banner entirely.

5. No first-10-users plan. Sunday evening arrives, the product is live, and you realise you don't actually know any humans to send it to. That's a Tuesday problem for someone who hasn't read this article. For you, you've had 10 names in a text file since Thursday. Send the messages.

Day 7 ship decision: keep, reshape, or kill

Monday through the following Sunday, you're not building anything new. You're watching how 10 real UK users interact with your thing.

By day 7 you'll know one of three things:

• Keep. Three or more people used it more than once. One asked when you're charging. Build the next feature, open a proper billing plan, and go find user 11.
• Reshape. People signed up, but nobody used the feature twice. The problem is real but the solution is wrong. Talk to the users who signed up. Build version two next weekend with a different angle.
• Kill. Crickets. Nobody cared. That's fine. The whole stack cost you under GBP 15 and you've learned which idea not to build. Pick the next one off your list and go again.

Weekend micro-SaaS only works as a habit, not a lottery ticket. The indie-hackers who are making GBP 5,000 a month in 2026 have usually shipped five things to ship the one that stuck. You're not swinging for a unicorn — you're building a portfolio of small bets, each of which cost you a weekend and a tenner.

Open your laptop. Friday's coming.

Want this week's UK builder-ready ideas?

[[Latest free report -> /reports]] — IdeaStack's weekly shortlist of UK-first micro-SaaS opportunities, scored for keyword demand, competitor gaps, and weekend-buildability.