Building Client Sites with Fabrement: An Honest Account
I’m a developer at 7Fridays, and I should say the obvious thing first: we built Fabrement. So this isn’t an independent review, and if you want one you should go and find one.
What it is instead is the account of someone who has used it on paid client work for since June, has the same deadlines as you, and gets to file bug reports directly at the people who wrote it. I’ve included the things it still does badly, because those are the parts I’d want to read.
Two ways to build a WordPress site, and why both were bad
For years there were two honest options, and I’ve done both.
Build it properly : Gutenberg, ACF in a fully custom theme. Clean, fast, maintainable. It also took weeks, not because the hard parts were hard, but because the routine parts were endless: moving designs out of Figma section by section, hand-building blocks, wiring responsiveness, then a week of adjustment rounds.
Or build it quickly : a premade multipurpose theme, a page builder on top, and a plugin for every small thing. Live in days. Then you pay for it: a maintenance burden that grows every year, updates that break layouts, and performance that collapses under the weight of code the page never needed.
Most of the industry chose the second, and I understand why. It won on the only metric a client can see at the start — how fast something appears. That’s what earned WordPress its reputation for being cluttered and slow. Not WordPress itself, and not custom development. The reputation came from what the shortcut left behind.
But the trade-off was real. Speed cost you quality, quality cost you time, and you picked your poison.
What changed is that AI removed the time cost from the clean path. When layout, blocks and responsiveness take twenty minutes instead of a day, the reason to accept a bloated theme and thirty plugins disappears. The shortcut made sense when the alternative was three weeks. It doesn’t when the alternative is an afternoon.
I’d still reach for a page builder on a five-page brochure site nobody will touch again. For anything a business depends on and has to maintain for years, the calculus has flipped.
Why I was sceptical anyway
Knowing all that didn’t make me an easy sell. I’ve watched a lot of tools promise to remove the boring part of WordPress development, and most of them removed the boring part and added a worse one, a proprietary runtime, a lock-in, a layer you spend the next two years fighting. That’s the same mistake as the page builder, wearing a newer badge.
So when we started building this, my concern was never whether AI could generate a hero section. It was whether what came out the other end would be something I’d be willing to hand to a client and still support in eighteen months.
What changed in my workflow
The shape of the work changed more than the speed did:
Describe the result → the agent builds it → I review and refine.
I’m still there in step three, and that matters. But I start from something that exists instead of an empty file, and reviewing catches in minutes what used to take a day to produce.
What I lean on most:
- Figma sync : pulling a design in without transferring sections by hand. Biggest single time saver, and the reason most of our projects touch it at all.
- Blocks from prompts or screenshots : for the standard furniture: heroes, feature grids, testimonials, pricing tables.
- Design system enforcement : colours, typography and spacing applied consistently instead of drifting across thirty sections.
- Native Gutenberg output : clean semantic markup that loads only on the pages using it, which is why our Core Web Vitals numbers survived the switch.
That last one answers my original scepticism. Page builders solve this problem by shipping a heavy runtime to every visitor forever. Fabrement treats WordPress core as a PHP framework and generates only the code a page actually uses, so what reaches the browser is the markup for that page and nothing else.
I use it for custom blocks, single post and archive templates, custom post types, reusable dynamic sections, and WooCommerce product pages.
What changed recently: the MCP connection
The newest addition is a Model Context Protocol connection, and it’s the biggest shift since we started.
MCP is an open standard for letting an AI agent talk to external systems through one common interface. In practice it means the agent isn’t limited to generating markup any more, it can build pages, create posts and fill the content, and wire up forms on the site as part of the same flow.
That matters more than it sounds. Generating a beautiful static section was never the hard part of client work, make it work with dynamic data was. Pulling that into the same workflow removes the step where a design hand-off becomes a two-week integration ticket.
If you want the longer technical version, there’s a full write-up on the Fabrement blog: .
Where it still falls short
The section I’d actually want to read.
Single post templates need too much hand-holding. It doesn’t yet reliably detect the standard furniture of a post – title, featured image, gallery, video, author, categories – so I end up specifying those explicitly more often than I should.
Typography control is thinner than I want. Multiple fonts across a design still means dropping into code. For a tool whose whole argument is design system consistency, that’s the wrong place to have a rough edge.
It is not a “no developer needed” tool. Someone still has to know what good WordPress architecture looks like, or you get a fast build of the wrong thing. My own rule: the content model with post types, taxonomies, how records relate gets decided by a person before the agent touches anything. Everything downstream of that is safe to delegate. That decision isn’t.
But the work that used to take several hours, sometimes a full day, now takes 20–30 minutes plus review. Even when adjustments are needed, total build time drops sharply.
Would I recommend it?
I’m the wrong person to ask, which is why I put the disclosure at the top.
What I can tell you is that I use it on client work every week, that I’d notice immediately if it stopped working, and that the two weaknesses above are on the roadmap rather than being quietly ignored, I know, because I’m one of the people complaining about them internally.
If you build WordPress sites for a living, the fastest way to judge it is to point it at one real section you’ve built by hand often enough to know what good looks like. The plugin is free.
- Fabrement —
- Download the plugin —
- Tutorials —
Future of WordPress with AI
