How to Build an Agency Website Without Writing Code
Write down what your agency does, for whom, with proof, then let an AI builder generate the site and edit the draft until it sounds like you. On Zugo a one-page agency site is a build of about a minute, and a multi-page version with separate case studies takes a few minutes and 12 credits to start.
The hard part is not the build. It is that agency websites converge on the same six sections and the same three adjectives, and the visitor cannot tell any of them apart. This guide is mostly about avoiding that, because the tool will happily produce the average of every agency site ever written.
What separates an agency site that works from one that does not?
Specificity, and nothing else. A prospect lands with one question: have you solved my exact problem for a company my size? A site either answers it in ten seconds or gets closed.
That means the sections that matter are the ones most agencies treat as secondary. Case studies with numbers beat a services grid. A named client type beats "businesses of all sizes". A stated price floor beats a contact form, because the enquiries you lose were never going to buy.
The sections most agencies lead with, a value-proposition headline and a capability list, are the ones a visitor skims. Keep them short, put them above proof, and let the proof do the work.
There is a useful test for any sentence on the page. Cover the logo and read it. If it could belong to four other agencies you can name, it is filler, and filler is what makes a site forgettable rather than what makes it wrong. Cut it or make it specific.
What should the prompt say?
Name the niche, the proof, and the one action. Generic in, generic out, and an AI builder is unusually literal about this.
Build a website for a five-person B2B content marketing agency working
with SaaS companies between 10 and 100 employees.
Pages:
1. Home: one-line positioning, three case study cards with a result
number on each, a four-step process, one call to action.
2. Case studies: index plus a template with client, problem, what we
did, what changed, and a quote.
3. Services: three engagements with what is included and a from-price.
4. About: five team photos with names and roles, how we work.
5. Contact: form with company, size, budget band, and what they need.
Tone: direct, specific, no superlatives, no stock photography.
Two constraints in that prompt are doing most of the work. "A result number on each" forces the case study cards to carry evidence instead of adjectives. "No superlatives" prevents the draft from arriving full of "world-class" and "innovative", which you would otherwise spend three edits removing.
If a blank field is not where you want to start, Zugo ships 25 templates, and forking one that already has the structure you want means your first move is replacing text rather than describing layout.
How should a case study be structured?
Four blocks, in this order, every time. The order matters more than the writing, because a reader scanning five case studies is comparing shapes.
| Block | What goes in it | Common mistake |
|---|---|---|
| Client and context | Industry, size, what they sell | Naming the client and nothing else |
| The problem | The situation before you arrived | Describing your service instead |
| What you did | Specific actions, in sequence | A list of deliverables with no reasoning |
| What changed | A number, a time frame, a quote | Vague "improved performance" |
If you cannot put a number in the fourth block, say so and use a qualitative outcome instead. An honest "the client stopped missing publication deadlines for the first time in a year" is more persuasive than a percentage nobody believes.
Three good case studies outperform twelve thin ones. If you have only one, publish one, and make it excellent. A site that shows a single deep engagement reads as a specialist; a site with twelve one-paragraph logos reads as a directory.
What does an agency site cost to build?
Pricing is per action rather than per page. A single-page site, app, or game is 6 credits. A multi-page platform is 12 credits covering the first three pages, with 3 for each page after that, and every edit is 3.
A five-page agency site is therefore 12 plus two extra pages, or 18 credits before you change anything. Add fifteen edits, which is realistic for copy you care about, and you are at 63 credits. Free comes with 5 credits, Pro is $25 a month for 200 credits, and Business is $99 a month.
Case study pages are worth mentioning separately. Each one you add later is 3 credits as a page, plus the edits to get it right. If you expect to publish a new case study monthly, the cheaper structure is to store them in a Supabase table and let the site read from it, so a new study is a database row rather than a build.
How long until it is live and on your own domain?
Generation is about a minute for a single page and a few minutes for a multi-page platform. The edits take an evening, and they are the real work: rewriting the positioning line, cutting the adjectives, getting each case study to fit the shape above.
Publishing is immediate. The project gets a your-agency.zugo.run address the moment you publish, which is enough to send a link to a colleague for review. A custom domain adds a DNS step and its own wait, and it is worth connecting yours before you start sharing the link widely rather than after.
Every build is booted in a sandbox before it reaches you, and a build that fails to open is reported as a failure rather than delivered as a blank screen. That lowers the risk of publishing something dead. It does not tell you whether your positioning line is any good, which remains the part only you can judge.
Which connectors does an agency site actually use?
Three, usually, and none of them on day one.
Resend delivers the contact form to your inbox. Until it is connected, a form renders and submits into nothing, which is the quietest way to lose a lead. Test your own form before you publish.
Supabase gives you a database, a login, and file storage. For agencies it earns its place in two situations: case studies that grow monthly, and a simple client login for sharing deliverables.
Google Analytics answers which case study gets read to the end. On an agency site that is the only feedback loop you have about whether the proof is landing.
GitHub export and the Vercel connector matter for a different reason. They mean the site is portable: source code in your repository, hosting on your account if you want it. An agency selling digital work should probably own its own site outright, and that path is open by default.
Where does an AI builder stop for an agency?
Four limits worth naming, because an agency's own site is a credibility artefact and the failure modes are visible.
It will not invent your positioning. The site can only be as specific as the prompt, and the prompt can only be as specific as your thinking. An agency that has not decided who it serves will get a beautiful page saying nothing.
Highly custom interaction takes iteration. An unusual scroll-driven case study, a bespoke animation, or an interactive pricing configurator is reachable through several rounds of edits, not one description.
It is a website, not a proposal system. Pipelines, contracts, and delivery workflow are separate tools. A simple CRM is a reasonable second project, but do not expect it to fall out of the same prompt.
Complex products need a team. Zugo does not replace a development team when a project grows into a real application. When it does, GitHub export gives you an ordinary repository, and the project belongs to you.
For a small agency that needs a specific, credible site published this week rather than next quarter, none of that is a blocker. If your work is closer to individual practice than team delivery, the same discipline applies at a smaller scale: see building a portfolio site. Otherwise, write the positioning line first, then describe the site around it at zugo.dev.