Our method: Blue Ocean positioning and StoryBrand messaging - how we build sites that win
Expertise - our build platform

We build on Claude Code.

Most agencies rent you a platform. We write your site as plain, fast code with Claude Code - an AI engineer working inside your codebase - so it’s researched properly, launches in days, scores 100 across Google Lighthouse, and never holds your content hostage.

Anthropic Claude Partner ProgramAnthropic Claude Partner
The whole idea, in one line

You can outsource the thinking. You can never outsource the understanding.

The typing, the boilerplate, the thousand small engineering decisions - hand those to the machine. What it can’t hold is the understanding of your business, your buyers and what actually matters. That stays with people: you, and a senior person here who has done this many times. AI makes the building cheap and fast. Judgement is still the job - and that’s exactly why this approach works rather than producing generic slop.

This is our recommendation, and we think it’s where web development is going. Not for everyone - but for most B2B companies, it’s already the better deal.

The short version

Live in days, not months

AI-speed engineering compresses the build. A site that would take a traditional agency a quarter is designed, built and deployed in a fraction of the time - without cutting the strategy that makes it work.

Lighthouse 100s, every build

Plain, fast code with nothing rotting underneath. Performance, accessibility, best practices and SEO - full marks, because we can reach and fix everything Google measures.

Changes the same day

The expensive part of a website is every change after launch. Here you describe what you want and it ships in hours, so the site keeps improving instead of going stale.

Owned by you, forever

It’s plain code in a repository in your name. No platform subscription, no export wall, no agency lock-in. If you ever leave, you leave with everything.

What “built on Claude Code” means.

More agencies are borrowing the phrase, so here is precisely what we mean by it - and what we don’t.

When we say a site is built on Claude Code, we mean every page, every component and every line of it exists as plain, typed code in a Git repository, written by Anthropic’s agentic coding tool under the direction of a senior person here. Claude Code isn’t a website builder and it isn’t a chat window that spits out snippets. It’s an AI engineer that works inside the codebase itself: it reads the project, writes and edits files, runs the tests, fixes what fails and commits the result. Direction stays human; execution is automated.

What it doesn’t mean is that we typed a prompt and a website fell out. Prompt-to-site tools exist, and they produce what you’d expect: something generic that nobody researched and nobody owns. A Claude Code build runs the other way round. The AI is handed real keyword and competitor data, a design you’ve signed off and a content plan, and it executes against them the way a good engineer executes a good brief. The build is fast because the typing is automated. It’s good because the thinking never was.

And if you’ve been searching for a Claude Code agency, or wondering whether you need to hire Claude Code developers to get a site built this way, the practical answer is simpler: you hire the output. We carry the tooling and the subscriptions; you get the finished asset - a repository in your name, deployed on your accounts, with websites from £4,500 on our pricing page. The site you’re reading now was built exactly this way, which is about as first-hand as evidence gets.

Not 85. Not 90.

Google Lighthouse scores every site on performance, accessibility, best practices and SEO. Template builds plateau in the 70s and 80s because nobody can reach the code underneath. Ours is plain code - so every flag Lighthouse has ever raised, we hand to the AI and it gets fixed. All four scores, green, full marks.

100PERFORMANCE
100ACCESSIBILITY
100BEST PRACTICES
100SEO

Why we build
this way.

The usual choice is framed as “do you want editor access?” - yes means a website builder, no means custom code. We think that’s the wrong question.

The real question is why you want editor access. Almost always the answer is “so my team can change the site after launch.” But you don’t need a team for that anymore - you describe the change and the AI makes it, usually live the same day. So it isn’t “editor or not”; it’s whether you keep paying people to push pixels, or just say what you want and have it shipped. That’s why we build on Claude Code: plain, fast code, changed by asking.

If hands-on, drag-and-drop editing is genuinely the priority for your team, we still build on Duda, and we’ll tell you so on the call. For everyone else, this is the better deal.

Against the usual ways.

Web Hero × Claude CodeSite buildersTraditional agency CMS
First version liveDaysWeeksMonths
Lighthouse scores100 / 100 / 100 / 10070–90, capped by the platformDepends who built it
Adding a page of contentPaste in the doc - system does the restRe-layout every block by handRaise a ticket, wait a sprint
When something breaksDescribe it; the AI fixes itHope a plugin existsDay rates and a queue
Who owns itYou - plain code, your repoThe platformSometimes you

What that changes.

Your design system, applied everywhere

We design the system once - type, colour, spacing, components. From then on every page, old or new, comes out looking like it was art-directed, because it was.

Content is effortless - and human

New page? Paste in a document your team wrote. The design system lays it out, and the schemas, metadata and links are generated on the way through. Nothing has to be written by AI - it just gets published by it.

Every problem is solvable

On a platform, you wait for a plugin. In plain code, you describe the problem and the AI fixes it - every bug, every audit flag, every “could the site just…?” we’ve ever been given, solved.

Found by search - and by AI

Structured data, semantic markup and instant load times are baked into every build. That’s what ranks in Google, and it’s what gets quoted when buyers ask an AI assistant who to shortlist.

How we
build it.

Research first, design signed off, then built and shipped. The AI accelerates the work; the judgement at every stage stays human.

01

Research

We pull real demand data with the DataForSEO API - keywords, search volumes, the SERP your buyers actually see - and layer in AEO and GEO context from Peec AI, so the site is built to be found in Google and cited by AI assistants from day one, not retrofitted later.

02

Design

We gather references you like, agree the direction, and design the key pages against your real competitors. You sign off the look and feel before a line of the production site is written - no surprises at launch.

03

Wireframe & content

Pages are wireframed around the buyer journey, then populated: your words where you have them, AI-assisted drafting where you don’t, always edited by a human who knows the subject. Structure first, polish second.

04

Build

Claude Code writes the site as clean, typed, accessible code on a modern stack. The same senior person directs the design and the build, so nothing is lost in a hand-off.

05

Deploy

Shipped through GitHub and deployed on Vercel-class infrastructure - versioned, previewable on every change, rolled out with zero downtime. Your repo, your hosting accounts, your control.

Research, not
guesswork.

Before a single pixel is designed, the AI is handed real data: where you sit today, where your competitors sit, and where you need to be - in Google and in AI answers.

DataForSEO

SEO context, from the API

We pull live demand through the DataForSEO API - keyword difficulty, search volumes, the real SERP your buyers see and exactly where your competitors rank - and feed all of it to the AI as context. The build targets winnable terms from day one instead of guessing.

Peec AI

AEO & GEO context

We layer in Peec AI’s API for Answer- and Generative-Engine Optimisation: how your category is being answered by AI assistants, what gets cited, and what your site has to do to become the source those answers quote.

Google Lighthouse

Measured, then proven

Every build is held to Google Lighthouse - performance, accessibility, best practices and SEO - and we tune to 100s, not high-80s, because in plain code we can reach and fix everything it flags.

Together these build a picture of where your business actually sits - its sitemap, its competitors, its gaps in search and in AI answers - and where it needs to get to. Only once that’s clear do we move to design. And after launch, the same APIs keep feeding the work: we read the data and keep improving your position month after month, without burning time on design and content the AI now handles. If you want the deeper story on how AI assistants decide what to cite - and how we build sites to be the source they quote - we’ve written it up in our guide to AEO and GEO.

The stack we build on.

Production-grade services, wired in as standard. You get a real application, not a brochure with a contact form.

Framework
Build tooling
Design system
3D & WebGL
GPU graphics
Markup
Styling
Language
Vector graphics
Database & auth
Data
Realtime & auth
Cache & queues
Hosting & deploys
Source control
Error monitoring
AikidoSecurity scanning
Quality audits
Transactional email
Notifications
Messaging
Content & ops
Content authoring
Documents
Payments
MolliePayments · EU
Product analytics
Automation
Search data API
AEO & GEO data
API marketplace
Structured data
Bot protection
AI models
AI models
AI models
AI models
Shipped, not staged

Proof, not promises.

Anyone can describe a workflow. These are live Claude Code builds - client sites and our own products - each shipped as plain code and owned outright by the people it was built for. Where a number appears, it’s the figure reported from the engagement, not a projection.

Client build · 3.4x demo requests, reported

Cademi

A multilingual training platform, rebranded and rebuilt around the demo funnel - full translation structure, SEO rebuilt from the foundations. Demo requests multiplied 3.4x after relaunch, reported from the engagement.

Read the case study

Our own product · end-to-end build

Avago

An AI website builder powered by Duda’s API. We built the marketing site, the API engineering, the account portal and the brand end to end on Claude Code. When it’s our own product on the line, this is still how we build.

Read the case study

Our own product · DEFRA integration

LoadSnap

A UK waste-compliance SaaS built to integrate with DEFRA Digital Waste Tracking, with a companion app and brand. Claude Code carried the software as well as the site.

Read the case study

Client build

SimpleFuel

Bulk fuel, fuel cards and telematics for fleet buyers - three product lines on one site without collapsing into commodity jargon, each with its own route to a qualified enquiry.

Read the case study

Client build

AMG Rib & Repair

RIB and inflatable-boat repair specialists. A fast, mobile-first site that turns research into qualified workshop enquiries - proof this approach works for specialist trades, not just SaaS.

Read the case study

Built with Coptrz

Drone Rules

A UK drone-regulation resource structured around the questions operators actually search, so both Google and AI assistants can quote it.

Read the case study

Each case study carries the detail: what was built, what changed, and how any number is verified. Where a figure hasn’t been signed off, we say so there too - that’s the standard the whole site is held to.

A site that keeps moving.

Launch is the start, not the finish. Here’s what living with a Claude Code site is like.

Update by asking

Need a new page, a price change, a fresh case study? Describe it in plain words. There’s no page builder to wrestle and no sprint to wait for - the change is made and pushed live, usually the same day.

SEO that compounds

We keep analysing and optimising after launch: search and AI-citation performance reviewed continuously, technical fixes and content shipped as the data tells us to, so the site gets stronger month on month.

Look and feel, when you want it

Design isn’t the point of this approach - speed, ownership and findability are - but it’s fully customisable. Want it to look exactly a certain way? We art-direct it. Want it clean and out of the way? That too.

Making a change, after launch

There’s no admin panel and no ticket queue. You send us the change in plain words; we make it and it goes live, usually the same day.

You
Can we add our new case study to the homepage, and update the team page with two new starters?
Web Hero
Done - both changes are live. We’ve linked the case study from your services page too.
shipped the same day · nothing for you to update
Already have a site?

Migrate it without losing your rankings.

Most of the sites we build on Claude Code aren’t first websites - they’re migrations off WordPress, Wix, Squarespace, Webflow, Framer, Duda or an old hand-coded build. Rankings don’t have to be the casualty. Every migration here runs the same four checks, in order: a full URL map, one-to-one 301 redirects, metadata and structured-data carry-over, and post-launch verification - and the last check happens after launch, not before.

The platform-by-platform detail lives on our Claude Code Migrations hub. And if you’re planning a move yourself - with us or without us - our website migration checklist walks the whole process step by step.

The honest bit

“So am I committing to pay you forever?”.

It’s the natural worry, and it comes from the old model - where every change meant a human building it step by step, drag-and-drop, a day or two of work billed at £350–£500 a time. That’s gone. A new page or a price change is a prompt; it ships in about thirty minutes, often live instantly. The whole infrastructure of building has changed: it isn’t about time anymore, it’s about goals. You tell us the goal, the change happens, and it costs a fraction of what it used to.

And to be straight: this isn’t for everyone. There are brilliant agencies built entirely around human designers, and some businesses want exactly that - which is completely fine. If that’s you, go and find one, and we’ll tell you so on the call. We’re not the right fit for everybody. For most B2B companies, we are.

Free guide

Why Build on Claude Code.

The short version, as a PDF you can keep or share with whoever signs off the website - why this approach, what it changes, and what moving an existing site over involves.

  • Why plain, AI-written code beats renting a platform
  • How a Claude Code build hits Lighthouse 100s
  • What “change it by asking” looks like after launch
  • What migrating an existing site actually involves

Get the guide

No spam, no sales sequence. The download starts as soon as you submit.

Where this is heading.

Blogs without the busywork

Insight and blog content integrates cleanly and doesn’t need a heavy CMS or a constant publishing treadmill to perform. Publish when you have something worth saying; the structure does the SEO work.

An MCP for your site

We’re building a Model Context Protocol server so AI tools - including ours - can talk to your website directly: read it, update it, answer questions about it. Your site becomes something you can simply instruct.

Subscription API access

Ongoing, subscription-based access to the AI build pipeline, so changes and new pages keep flowing without a new quote every time. Capacity when you need it, paused when you don’t.

Background

What is Claude Code?

Claude Code is an agentic coding tool from Anthropic, the company behind the Claude family of AI models. It runs in a developer’s terminal and works directly inside a codebase - reading files, writing and editing code, running commands, and using tools - rather than just suggesting snippets in a chat window.

Anthropic introduced it in 2025, first as a research preview alongside a new generation of Claude models and then more widely as the models grew capable enough to carry real engineering work end to end. It quickly became one of the most-used ways professional developers actually build with AI, precisely because it operates on a real project the way an engineer does.

For a web studio that is the unlock. Earlier AI tools could autocomplete a function; Claude Code can take a researched brief and a design and build the site - typed, tested, accessible - then keep changing it on request. Paired with senior judgement on strategy, design and what’s worth building, it’s how we ship sites in days that a template could never match and a traditional agency would quote months for. It’s the engine; the understanding is still ours and yours.

It’s worth being clear about what Claude Code is not. It isn’t a website builder: there’s no theme gallery, no drag-and-drop canvas, no hosting plan attached. And it isn’t ChatGPT: a chat assistant answers in text and leaves you to do something with it, while Claude Code acts on a real project directly. If you’re weighing it against the other names you’ve heard - Cursor, Codex, Copilot - we’ve written an honest comparison from production experience: Claude Code vs Cursor, Codex and Copilot.

The tool itself is the cheap part. Subscriptions run from roughly $20 a month (checked July 2026 - Anthropic’s prices move), and we carry that cost, so it never lands on you. What the tool’s price does explain is why the economics of building changed so sharply, and we unpack that - plans, API rates and what a professionally built Claude Code site costs in the UK - in Claude Code pricing UK.

For years, building a website meant choosing a builder - Webflow, Framer, WordPress, Duda. We built on all of them, and we still build on Duda where hands-on editor access is genuinely the priority. Claude Code is the step beyond: the same speed of change, without the platform ceiling, the subscription or the lock-in.

This is the short version. The full plain-English guide - what Claude Code is, what “agentic” actually means, and the questions to ask any agency claiming to build with it - is at What is Claude Code? A guide for business buyers.

Fair questions.

Claude Code is an agentic coding tool from Anthropic, the company behind the Claude AI models. It works inside a real codebase - reading files, writing and editing code, running tests and shipping committed changes - rather than suggesting snippets in a chat window. We use it as the engineering engine on every build, directed by a senior human here. If you want the full plain-English version, our guide at /built-on-claude-code/what-is-claude-code goes deeper.

Related