Most tutorials show you the retail way: use Claude to design a site, then push it into WordPress and edit it by clicking. It works — for one site, for a hobby. But if you want to run this as an agency — charging for SEO and monthly website updates across many clients — the developer tool, Claude Code, is a different machine entirely. It's the difference between doing a job and owning a system. Here's exactly where they split, and why it decides whether your agency can scale.
Every future edit: click and drag inside the site's admin. Nice at first — but it never gets faster, and it's per-site, forever.
Every future edit: describe it, Claude Code changes the code, one command deploys. Tracked, reversible, and it scales across every client you have.
People say “I'll use Claude to build the site” as if there's one product. There are really two ways to put Claude to work, and they lead to very different businesses.
You design in Claude Design (the visual tool that turns a sentence into a good-looking site), then use the chat app plus a connector plugin to push that design into WordPress, where you finish it by hand in a page builder like Elementor.
Claude here is a helpful assistant bolted onto a website admin. It can nudge things, but the real site lives inside WordPress, and you are the one clicking.
Claude Code is Claude working as an actual software developer: it opens your project's real files, writes and edits the code, runs commands, publishes the site, and can set up things that run on their own afterward.
You don't click around an admin — you describe what you want and it does the engineering. This is the same class of tool professional developers use to ship real software.
This is the single distinction that decides whether your agency can take on the 10th client as easily as the 1st. It's about where the true copy of each site is kept — and who, or what, can change it cleanly.
The real version of the site is a big tangle inside WordPress's database, in the page builder's private format. Two hands edit the same tangle: you (clicking) and the AI (through the bridge). The moment you drag one margin by hand, the AI's picture of the site is stale — and there's no clean history to undo a mistake.
For one hobby site that's fine. Across a dozen client sites, it's a dozen fragile snowflakes that only you can touch.
The real version is a set of code files with a complete, saved history — like a document's revision history, but for the entire website. There is one true version. Anything can be undone, copied, or automated, because software can read and change code perfectly.
That's what makes a second, tenth, or hundredth site a copy-and-adjust job instead of a from-scratch job.
Why an agency should care: the retail path sells a site — a thing you hand over and hope doesn't break. The developer path builds an asset you keep improving — versioned, reproducible, and safe to change every month. That difference is the difference between a freelancer and an agency.
These come straight from the popular WordPress-bridge tutorial. They're fine once. Now multiply them by every client you want to serve.
Retail numbers are observed from the real tutorial. The developer side is qualitative on purpose — it depends on your own setup, and it would be dishonest to invent a benchmark.
Read down the left for what's being compared, then across. Clay is the retail WordPress-bridge way; blue is Claude Code, the developer way.
| What matters to an agency | RetailClaude → WordPress, by hand | DeveloperClaude Code, real code |
|---|---|---|
| Time to launch site #2, #3, #10 | Same hour of clicking every single time — the work never compounds. | Reuse your setup; new sites are copy-and-adjust, so each one is faster than the last. |
| Doing a monthly update | Log into each site, hunt for the element, drag it, hope nothing else shifts. | Describe the change; Claude Code edits the code and deploys. Same skill for every client. |
| If an edit breaks something | No clean site-wide undo — you fix it by hand, from memory. | Full history: roll back to the exact previous version in seconds. |
| Site speed (an SEO ranking factor) | Heavy — runs software on every visit, leans on caching plugins, page builders add bulk. | Very fast — pages are pre-built and served from servers near each visitor. |
| SEO control | Depends on plugins layered over cluttered page-builder markup you don't fully control. | Full control of the clean, lean details search engines actually read. |
| Making 50 local landing pages | Build each one by hand — practically impossible at scale. | Generate them from a list automatically (see the SEO section below). |
| Automation after launch | Limited on basic hosting — usually more plugins or manual work. | Built for it: auto images, uptime checks, lead alerts, scheduled reports. |
| Who can touch the site | The client can log into WordPress and break it between your invoices. | You own the asset; hand the client a small, safe editor for just their words. |
| What you're really selling | A one-time site build. | A managed, monitored, continuously-improved asset — i.e. a retainer. |
| Skill floor | Low — beginner-friendly clicking. Its one real advantage. | Higher to start — but Claude Code does the hard parts, and it pays back fast. |
↔ On a phone, scroll the table sideways to see both columns.
You said the agency offers SEO and website updates. Those are exactly the two jobs where the retail and developer paths diverge the most — so let's put each one under the microscope.
SEO is partly about the words on a page and partly about the invisible plumbing: how fast the site loads, how cleanly it's built, and how many relevant pages exist for people to find. This is where the developer path isn't a little better — it's a different category.
You install an SEO plugin on top of page-builder markup that's bulky by nature. Speed depends on caching plugins fighting the weight. Every extra landing page is a page you build by hand — so “a page for each city we serve” quietly never happens.
Clean, fast pages by default, with full control of the technical SEO details. And the killer move: from a simple list, Claude Code can generate hundreds of pages automatically — e.g. a tailored page for every service × city combination — the single biggest lever in local SEO.
The compounding advantage: “Car Accident Lawyer in Tacoma,” “…in Everett,” “…in Olympia” — hundreds of pages built from data in one pass. Impossible to hand-build in a page builder; routine on the developer path. This alone can be your whole pitch.
Updates are where an agency actually makes its living: the retainer. So the real question isn't “can I make a change?” — it's “how much of my day does one change eat, times every client?”
Each request means logging into that client's admin, finding the element, dragging it, and re-checking the layout didn't shift. The effort is the same every month and doesn't shrink — and it caps how many clients one person can carry.
You tell Claude Code “change the hero headline and add a testimonial,” it edits the code and ships it — with a saved record you can undo. The same short workflow works for client #1 and client #30, so your capacity actually grows.
The trap to avoid: the tutorial pitches “editing in the page builder costs no AI tokens” as the win. For an agency it's the opposite — it means every edit is your manual labor forever, and it silently cuts the AI out of the loop. Cheap per edit, expensive per month.
Same starting design, two completely different products to sell — and two different ceilings on what your agency can become.
A one-time build the client owns and can log into. Every future change is your manual labor or theirs. Great for a favor or a side gig.
Ceiling: a freelancer trading hours for one-off builds. More clients = proportionally more clicking.
A fast site you build and keep sharpening — new SEO pages, monthly updates, automatic monitoring and lead alerts — billed as an ongoing service.
Ceiling: a real agency with recurring revenue and a repeatable system. More clients = a bigger, tidier machine.
This isn't a hit piece on WordPress. It's the world's most popular website software for good reasons — and there are real cases where the retail path wins.
Claude + WordPress + a page builder
Claude Design → Claude Code
You don't need a computer-science degree — you need to let each tool do its job and build a small repeatable routine.
Have Claude Design build a reusable style kit (colors, fonts, spacing, buttons) before any site. Everything you make afterward matches automatically — that's the seed of a repeatable look across clients.
Use Claude Design only for the visual first draft. Don't try to make it run the live site — that's where the retail path bogs down. Export the design as the starting point for Claude Code.
Claude Code turns the design into clean code, wires up SEO, optimizes images automatically, and publishes to fast hosting. You describe; it engineers.
Nail one client site end to end. Your second client isn't a fresh hour of clicking — it's a copy of a system you adjust. That reuse is the agency.
Auto-generate location and service landing pages from a list. This is the offer no page-builder shop can match, and it's the easiest thing to charge a premium for.
Package it as ongoing: monthly updates, new SEO pages, monitoring and lead alerts. Recurring revenue is what turns a string of gigs into a business.