We knew this question was coming. It comes from a reasonable place. People have spent years using website builders that charge nothing upfront — pick a template, drag some blocks around, publish for free. So when Ullbek asks you to spend credits to build, it's fair to ask: what is that, exactly? What am I actually paying for?
The honest answer is more interesting than "server costs." It gets to the heart of what Ullbek is — and why it's fundamentally different from every drag-and-drop tool you've used before.
Drag-and-drop is free because nothing is being made
Here's the thing about template builders: when you use them, the tool isn't creating anything. It's rearranging. The HTML exists already — locked inside a template. The CSS exists already — baked into preset design tokens. You're moving pre-made pieces around a pre-made grid. The software just records your choices.
That's why it can be free. There's no act of creation happening on your behalf. There's no intelligence being applied to your specific brief. The tool doesn't know anything about your business, your aesthetic, or your ambitions — and it doesn't need to, because the template has already made all those decisions for you.
That's the real distinction. Every time you send a message to Ullbek's agent, real code is being written. Not retrieved. Not assembled from blocks. Written — by an AI model that has read your brief, understood your intent, and is producing original HTML, CSS and JavaScript tailored to what you described. That process has a cost. The credit is how you pay for it.
Think of it as paying a developer — not a subscription
The closest real-world analogy isn't "paying for software." It's paying a developer for their time and output.
When you hire a developer to build a website, you're paying for their expertise, their labour, and the code they produce. The code is yours at the end. It runs anywhere. It doesn't disappear if you stop paying a monthly fee. You own the output of the work.
- Original code, written for your brief. Every build is a unique output — not a template with your logo swapped in. Real HTML, CSS and JS designed around what you described.
- A developer's capability, without a developer's price. The agent applies the same reasoning a skilled developer would — layout decisions, typography choices, responsive behaviour — at a fraction of what that expertise would cost by the hour.
- Code you own outright. Export it, host it anywhere, hand it to a developer to extend. It's yours. No lock-in. No "site disappears if you cancel."
- The infrastructure we built around the model. The agent doesn't just run a prompt. It reads browser errors, fixes them, understands design context, and maintains consistency across your site. That harness is what makes the output good — and building it took us months.
Why we use the best models — and why that matters to you
We made a deliberate decision early on: we would never compromise on the quality of the underlying AI model to save costs. The baseline model we work with is Claude Sonnet — one of the most capable reasoning models available. We don't apologise for that, and here's why you should care.
A weaker model doesn't just produce worse code. It produces code that requires far more back-and-forth to get right. You spend more messages correcting mistakes, more time explaining what you meant, more frustration watching the agent misunderstand your brief in the same way twice. The credits you "save" by using a cheaper model are paid back in time — your time — fighting to get an output that a better model would have produced on the first try.
Good models understand nuance. They interpret a brief the way a skilled creative professional would — reading between the lines, applying taste, making educated choices where the instructions are ambiguous. That's not a luxury. For a product where the entire interface is language, it's the product.
What we built around the model
The model is the engine. But the car matters too. A lot of what you pay for with a build credit is not the raw model inference — it's the system we built around it.
- Live browser verification. The agent doesn't just write code and send it. It checks the output in a real browser, reads console errors, and fixes them — automatically, before you ever see a broken page.
- Design context awareness. The agent understands your site's existing design language — your colours, fonts, spacing — and applies it consistently to everything new it builds, without you having to repeat yourself.
- Intent interpretation. "Make it warmer" is a real instruction our system understands. Translating that kind of qualitative feedback into specific, correct code decisions is a hard engineering problem. We've spent months on it.
- Publishing guidance. When you're ready to go live, the agent walks you through DNS, domain setup, and SSL — built into the same conversation. No switching tabs, no reading documentation.
None of that comes from the model alone. It comes from the product we've built around the model. That's what you're investing in when you spend a credit — and it's why we're confident the value is real.
A freelance developer charges hundreds of pounds per day. An agency charges thousands per project. A credit with Ullbek costs less than a coffee. The gap between what you're paying and what you'd pay for equivalent human expertise is, frankly, extraordinary. We intend to keep it that way.
Start with $1 in free credits
No card required. See exactly what a build credit produces — a real, original website built for your brief.