My production stack in 2026: Next.js, TypeScript, Postgres, Stripe, Clerk
The stack I ship every UK SaaS build on, why each piece earned its place, and the ones I keep getting asked about but refuse to use.
Founders ask me on every scope call what the stack is and why. So here is the opinionated answer in one place. This is the stack I use on every UK SaaS build I take on, from regulated pharmacy platforms to compliance dashboards to digital banking flows.
Next.js (App Router, TypeScript)
Server components, route handlers, server actions, edge runtime where it makes sense. The App Router is no longer the new thing, it is the default thing, and in 2026 there is no faster path from idea to production SaaS than Next.js with TypeScript.
If anyone is still pitching you a Create React App spa with a separate Express backend, run.
Postgres (via Neon or Supabase)
Boring, battle tested, will outlive every framework debate. I reach for Neon when I want serverless branching for previews, Supabase when I want auth and storage in the same dashboard. Either way you end up writing real SQL eventually, and Postgres rewards that.
MongoDB is fine if you genuinely have unstructured data. Most founders do not. They have relational data and they keep telling themselves it is not relational because that document store demo looked easy.
Stripe
Subscriptions, metered billing, Connect for marketplaces, customer portal, tax. Stripe is one of the few APIs that gets better every year. I have not had a single project where another payment provider was the right answer for a UK founder.
Clerk
Auth is the most over engineered thing in software. Clerk solves it in an afternoon. Social login, magic links, MFA, organisations, billing tie ins. You can self host the auth flow if you want, but most UK SaaS founders are paying themselves to write auth instead of building the actual product. Clerk gets that hour back every day.
Vercel
Edge, ISR, previews per branch, image optimisation, analytics. The price doubles past the hobby tier but the time saved is worth a multiple of that. If you are a UK founder funding the bill personally, I will move you onto a self managed Docker setup the moment the maths flips. Until then, Vercel.
Resend for email
Postmark and SendGrid both still work. I use Resend because the developer experience is a foregone conclusion and the deliverability has been clean across every UK domain I have run through it.
Claude or OpenAI for AI features
If the build needs AI baked in, Claude is my default for reasoning heavy workloads and structured outputs, OpenAI when I need image input or speech. Most founders ask for AI features for marketing reasons. I push back hard until we find one where the AI is actually load bearing in the product.
What I refuse to use
- Bubble or Webflow for anything that needs to scale past a landing page
- MongoDB for relational data masquerading as documents
- Tailwind UI starter templates pretending to be design
- Firebase past the prototype phase (vendor lock plus expensive at scale)
- Any auth provider that prices per seat at the MVP stage
Why this stack wins UK founders work
Every piece in here is hireable. If we part ways in eighteen months, you can hire the next Next.js developer in the UK off LinkedIn and they will understand the codebase the same day. That portability matters more than which framework had the cooler launch video.
Working on something?
High ticket dev work, built in six weeks.
Senior freelance Next.js engineer in the UK. Fixed price, fixed scope, one engineer with the keyboard. Sprints from two weeks, full Builds from £10,000.