How to ship a SaaS MVP in a weekend with Claude Code

Can you actually ship a SaaS product in a weekend? Not a landing page. Not a side project that "might eventually charge money." A real, working product with authentication, a database, and a payment flow -- live on the internet by Sunday evening.

The honest answer is yes. With caveats.

Claude Code has changed what is possible for a solo builder in a compressed time window. It is not magic -- it will not replace judgment, and it will not save you from a bad idea -- but it does collapse the gap between "I have a concept" and "I have a working thing people can actually use."

This guide gives you a practical framework for doing exactly that. We will use a concrete example -- an AI-powered feedback form analyser for small businesses -- and walk through every stage from Friday evening spec to Sunday evening deploy. You will get real prompts to use at each stage and an honest account of what will and will not be done by Sunday.

---

Key takeaways

• Friday evening is for spec and setup only -- do not write a single line of code until you have a clear brief
• Claude Code works best when you treat it like a senior developer: give it context, constraints, and clear acceptance criteria
• Supabase handles auth and database, Stripe handles billing, Vercel handles deployment -- this stack deploys in under an hour once your code is ready
• A V1 MVP means one core workflow, not a full product -- scope ruthlessly or you will not ship
• Sunday evening is for smoke testing and going live, not adding features -- features come in week two

---

The example: FeedbackLens

Throughout this guide we will build FeedbackLens -- a tool that lets small business owners paste in customer reviews or feedback, and get back a structured AI analysis: sentiment breakdown, recurring themes, suggested actions. Simple concept, real value, clear monetisation path.

It is a good weekend project because it has:

• A single, understandable core workflow
• No complex real-time requirements
• A natural subscription model (pay per month, analyse unlimited feedback)
• A clear customer (small business owners who get reviews but lack time to read them properly)

Your idea does not have to be this one. But it should be this simple.

---

Friday evening: spec and setup (3 hours)

Step 1: Write a one-page spec

Before you open Claude Code, write a spec. This is the most important thing you will do all weekend. A vague brief produces vague code.

Your spec should cover:

• What it does -- one sentence, written as a user benefit
• Who it is for -- be specific (not "small businesses", but "sole traders and small teams who get customer reviews")
• Core workflow -- the single thing the user does and what they get back
• Data model -- what gets stored and why
• Auth requirements -- does it need accounts? (usually yes if you are charging)
• Pricing -- free tier? trial? flat monthly?
• Out of scope -- explicitly list what you are NOT building this weekend

For FeedbackLens, the spec looks like this:

What it does: Analyses customer feedback using AI and returns a structured report with sentiment, themes, and actions.

Who it is for: UK sole traders and small business owners who collect reviews (Google, Trustpilot, direct) but do not have time to synthesise them.

Core workflow: User pastes feedback text into a form, clicks Analyse, gets a structured report back within 30 seconds.

Data model: Users, FeedbackSessions (raw input + AI output + timestamp), Subscription status.

Auth: Email/password via Supabase. No OAuth this weekend.

Pricing: 3 free analyses, then GBP9/month for unlimited.

Out of scope: CSV import, team accounts, custom branding, API access, integrations.

Step 2: Set up your project structure

Open your terminal and give Claude Code this brief:

You are setting up a new SaaS project called FeedbackLens. The stack is:
- Next.js 14 with App Router
- TypeScript
- Supabase (auth + PostgreSQL database)
- Stripe (subscription billing)
- Vercel (deployment)
- Tailwind CSS + shadcn/ui for UI

Create the project structure. I want:
- /app directory with App Router
- /components directory
- /lib directory for Supabase client, Stripe helpers, and OpenAI client
- /app/api directory for API routes
- Environment variable template (.env.local.example) with all required keys

Do not install any packages yet -- just scaffold the directory structure and create the env template.

This prompt is explicit about what to do and, critically, what not to do. Claude Code works better with clear boundaries.

Step 3: Configure your services

While Claude Code is scaffolding, open three browser tabs and set up:

1. Supabase -- create a new project, note your URL and anon key
2. Stripe -- create a new product (FeedbackLens Pro, GBP9/month), note your publishable and secret keys
3. OpenAI (or Anthropic) -- grab your API key for the analysis feature

Fill in your .env.local file. This feels mundane but doing it now means you will not be blocked mid-build on Saturday.

---

Saturday: build the core features (8-10 hours)

Saturday is execution day. Work in focused 90-minute blocks with breaks. Do not scope creep. If you think of a great feature at 3pm, add it to a Notion doc and keep building.

Block 1: Database schema and auth (90 minutes)

Give Claude Code this prompt:

Create the Supabase database schema for FeedbackLens. I need:

Tables:
- profiles (id references auth.users, email, created_at, subscription_status: 'free'|'pro', analyses_used: int default 0)
- feedback_sessions (id, user_id references profiles, raw_input: text, ai_output: jsonb, created_at)

Write the SQL migration file. Then create a Supabase client helper in /lib/supabase/client.ts (browser) and /lib/supabase/server.ts (server-side with cookies).

Also set up Row Level Security: users can only read and write their own data.

Once the schema is created, set up auth. Give Claude Code this prompt:

Implement email/password authentication using Supabase Auth. I need:
- /app/(auth)/login/page.tsx -- login form
- /app/(auth)/signup/page.tsx -- signup form
- /app/(auth)/auth/callback/route.ts -- Supabase auth callback handler
- Middleware to protect /dashboard routes (redirect to /login if not authenticated)
- A basic layout for auth pages

Use shadcn/ui components for the forms. Keep it clean and minimal -- no fancy animations.

Block 2: Core analysis feature (90 minutes)

This is the thing your product actually does. Give Claude Code this prompt:

Build the core analysis feature for FeedbackLens.

I need:
1. An API route at /app/api/analyse/route.ts that:
   - Accepts POST with { feedback_text: string }
   - Validates the user is authenticated and under their analysis limit (3 for free, unlimited for pro)
   - Calls OpenAI (gpt-4o) with this system prompt: "You are a business analyst. Analyse the following customer feedback and return structured JSON with: overall_sentiment (positive/neutral/negative), sentiment_score (0-100), themes (array of strings), positive_highlights (array of strings), areas_for_improvement (array of strings), recommended_actions (array of strings, max 3)"
   - Saves the session to the database
   - Returns the structured result
   - Increments analyses_used on the profile

2. The main dashboard page at /app/dashboard/page.tsx with:
   - A textarea for pasting feedback
   - An Analyse button that calls the API
   - A loading state
   - The results displayed clearly (sentiment badge, themes as chips, actions as a checklist)
   - A count showing "X of 3 free analyses used" for free users

This is the longest and most complex block. Claude Code may need a few back-and-forth corrections here. That is normal. When it gets something wrong, be specific:

The sentiment score is not showing correctly -- it is displaying as a decimal between 0 and 1 instead of a percentage. Fix that.

Specific corrections beat vague complaints.

Block 3: Stripe billing (90 minutes)

Payment is the most technically fiddly part but Claude Code handles it well if you are explicit. Use this prompt:

Implement Stripe subscription billing for FeedbackLens.

Setup:
- Stripe secret key is in STRIPE_SECRET_KEY
- Monthly plan price ID is in STRIPE_PRICE_ID (GBP9/month)
- Webhook secret in STRIPE_WEBHOOK_SECRET

I need:
1. /app/api/stripe/checkout/route.ts -- creates a Stripe Checkout Session for the pro plan, passes the user's email and Supabase user ID as metadata, redirects to /dashboard?upgraded=true on success
2. /app/api/stripe/webhooks/route.ts -- handles checkout.session.completed (update user subscription_status to 'pro' in Supabase) and customer.subscription.deleted (revert to 'free')
3. /app/api/stripe/portal/route.ts -- creates a Stripe Customer Portal session so users can manage/cancel their subscription
4. An upgrade prompt component shown when free users hit their 3-analysis limit
5. An "Upgrade to Pro" button in the dashboard header

Use Stripe's raw body parsing for the webhook route (important for signature verification).

The note about raw body parsing is important -- Claude Code sometimes forgets this and the webhook will fail silently. If your webhook is not working on Sunday, this is the first thing to check.

Block 4: Polish and edge cases (90 minutes)

By mid-Saturday afternoon you should have a working product. Spend the final block on:

Polish the FeedbackLens dashboard:
1. Add a /dashboard/history page that shows all past analysis sessions for the logged-in user, with date, a snippet of the input, and a link to view the full result
2. Add proper error handling throughout -- if the API fails, show a user-friendly message, not a raw error
3. Add a loading skeleton for the history page
4. Make sure the layout is responsive on mobile
5. Add a simple /pricing page that explains free vs pro, with an upgrade button
6. Set the page title and meta description for each page

Do not add any new features. Focus on making what we have solid.

---

Sunday morning: polish and deploy (4-5 hours)

Step 1: Environment and deployment prep

Create a Vercel project and connect your repository. Then give Claude Code this prompt:

Help me prepare FeedbackLens for production deployment on Vercel.

1. Create a vercel.json if needed for any custom config
2. Check for any hardcoded localhost references and replace with environment variables
3. Make sure all Stripe webhook handling is set to use the production webhook secret
4. Add a simple /app/page.tsx landing page (if we do not have one) with: headline, subheadline, what FeedbackLens does, pricing table (free vs GBP9/month pro), and a "Get started free" CTA
5. List all environment variables I need to set in Vercel

Do not add new features -- just make it production ready.

Step 2: Set environment variables in Vercel

Add all your environment variables in Vercel's dashboard. Do this before deploying -- missing env vars are the most common reason first deploys fail silently.

Step 3: Deploy and smoke test

Push to main, trigger a Vercel deploy, then work through this checklist:

1. Sign up with a new email -- does the confirmation flow work?
2. Log in -- does redirect to dashboard work?
3. Paste in some feedback and click Analyse -- does the result come back?
4. Does the analysis counter increment?
5. Hit the 3-analysis limit -- does the upgrade prompt appear?
6. Go through the Stripe checkout (use test card 4242 4242 4242 4242) -- does the account upgrade?
7. Check Stripe dashboard -- is the subscription recorded?
8. Log out and back in -- is the upgrade preserved?

If you hit issues, Claude Code can debug. Give it the error message and context:

The webhook is returning 400. Here is the Vercel function log: [paste log]. Here is my webhook handler: [paste code]. What is wrong?

Step 4: Set up the real Stripe webhook

In Stripe's dashboard, create a production webhook pointing to https://your-domain.vercel.app/api/stripe/webhooks. Add the signing secret to your Vercel environment variables and redeploy.

Step 5: Go live

Switch Stripe from test to live mode, update your environment variables with live keys, and redeploy. You now have a live SaaS product.

---

"But can you really ship in a weekend?"

Yes. With these caveats:

You will ship: A working product with real auth, a real core feature, real billing, and a real deploy. Something you can show people and charge money for.

You will not ship: A polished, production-hardened, edge-case-proof application. V1 is intentionally incomplete. The first customers are also your QA team.

The hardest bit is not the code. It is keeping scope tight. Every time you think "I should also add X", you are threatening your Sunday deploy. Write it down, build it in week two.

Claude Code is a multiplier, not a replacement for judgment. It will happily build the wrong thing efficiently if you give it a bad brief. The spec you write on Friday evening is the most important work of the weekend.

Expect bugs on Monday. That is fine. A live product with bugs is infinitely more useful than a perfect product that does not exist.

---

FAQs

Do I need to know how to code to use this framework? Some coding knowledge helps -- you need to understand what you are asking Claude Code to build and be able to read the output. But you do not need to be a senior developer. Many successful weekend builds have been done by people who code occasionally rather than professionally.

What if I want to use a different stack? The framework works with any stack. Supabase + Stripe + Vercel is a solid default because all three have free tiers, good documentation, and work well together. But you could swap in PlanetScale, Lemon Squeezy, Railway, or any other provider -- just adjust your prompts accordingly.

How much will this cost in API fees? For a weekend build, minimal. OpenAI's GPT-4o costs roughly $0.005 per analysis request. Supabase is free up to 500MB. Stripe charges 1.4% + 20p per transaction (UK cards). Vercel is free for hobby projects. Total cost for the weekend: probably under GBP5 in API fees.

What happens after launch? Week two is for fixing the bugs your first users find, adding the second most important feature on your list, and starting to drive traffic. The IdeaStack weekly report can help you spot where the genuine demand is in your category.

Can I sell the product I build over a weekend? Yes. There is no reason a weekend V1 cannot be a real commercial product. Keep the scope tight, price it honestly, and be transparent with early customers that it is V1. Most people appreciate honesty more than polish.

---

Ready to find your next build? Every Thursday, IdeaStack publishes a deeply researched UK business opportunity -- complete with keyword data, competitor analysis, and copy-paste builder prompts. Read the latest free report ->