The business card plus a panel, a catalog and an API contract — you publish the content yourself.
Someone who publishes their own content: blog, catalog, events, products.
Everything from the business card plus a Strapi panel with content types shaped around your domain, lists with filters and pagination, categories and tags as their own pages, and a public API contract.
The front generates its client from the contract and keeps it in the repository, so a change to the content model shows up as a change in the code — not as a surprise in production.
A complete company site: a design made for your subject, the copy, and a launch with certificate, email and analytics.
A marketplace theme is picked to match an industry. We pick the design to match the content. First we know what the site has to say and to whom, and only then does AI propose a visual direction from a base of 67 styles, 161 palettes and 57 font pairings. A human makes the call and writes it down as the visual canon in the repository — that file then outranks any tool suggestion, so the site does not drift as changes pile up.
No Figma mock-up stage and no "template number 7". The first pass of the layout takes hours, and the time normally spent clicking through mock-ups goes back into the content and the details.
What you getVisual direction, the first pass of layout and components, draft copy, images, the OG image, alt texts.
The direction decision and its record, semantics and accessibility, performance, structured data that matches what is on the page, a review of all copy, deployment and maintenance.
A site whose content you change yourself: blog, catalog, events, products. Strapi 5 for the panel, Next.js for the front.
Two separate repositories joined by an OpenAPI contract. The API publishes a static contract file, the front generates a client from it and keeps that client in its own repository.
It sounds like a detail and it decides how calm your Sundays are: the build needs no access to the API, so it is deterministic, and every change to the content model shows up as a change in the front-end code — not as a production outage.
What you getThe content model from a description of the domain, list and detail components, SEO fields, sample data migration.
Model boundaries (what is a relation and what is a component), permissions, query performance, avoiding n+1, caching and its invalidation.
Filters, state kept in the URL, pagination, full-text search with a synonym dictionary.
Filters by category, location and tags, filter state in the address (the link can be shared), pagination or load-more, full-text search and a synonym dictionary for the words visitors use differently than the editors.
A trap we know from our own outage: a catch-all filter route caches every URL permutation a crawler invents. Ours grew to 38 GB across 1.4 million files and took five services down. So a catalog here gets `noindex, follow` on filtered views by default, an allow-list of parameter values, and a limit on what may enter the cache at all.
What you getList, filter and search components, and a first synonym dictionary.
Cache and indexing boundaries, the parameter allow-list, and query performance as the catalog grows.
A REST API with a panel, a public OpenAPI contract, and a client generated from that contract on the consumer side.
Build-time validation checks that every path is in the contract, that every type carries all its fields, and that relations expand to a set depth without looping. On top of that: separate read and write tokens, a healthcheck, and upload size limits set deliberately per project.
What you getGenerating the client and the types, a first validation pass.
The shape of the contract, relation depth, token permissions and limits.
Metadata, canonicals, sitemap, internal linking — and the decision about what belongs in the index at all.
The most common mistake on AI-generated sites: everything is indexable. Thousands of filter URLs, the crawl budget burned, and the site still does not grow. Here filtered views get `noindex, follow` and a canonical to the clean path.
Internal linking is a feature, not editorial goodwill: "more in this category", "previous and next", "similar by tags" and category and tag hub pages are generated from the data.
What you getDraft meta descriptions, spotting missing metadata, generating OG images.
Deciding what belongs in the index, and making sure internal links come from the data rather than from an editor’s memory.
Describing content in the format search engines and AI models read. Here it is a requirement, not an add-on.
One graph per page. `Organization` and `WebSite` with a search action sit once in the layout, and page nodes link to them by `@id`. On top of that: breadcrumbs and the narrowest entity type for the page — `Store`, `BlogPosting`, `Event`, `Product` with an offer, `FAQPage`, `HowTo`, `Person`, `Book`, `VideoObject`.
The hard rule: structured data must match what is visible on the page. A language model will happily add a 4.9 rating from 87 reviews that do not exist. That is a straight path to a penalty, not to stars in the results.
What you getA first pass of the graph from the content model, and spotting gaps.
Matching the type to the content and making sure the graph claims nothing the page does not show.
Page copy, articles, category descriptions and meta descriptions — written to a written canon, not "generated".
One steady author for the whole site, the choice recorded in the repository. Someone who cares about the subject, not an "editorial team". Spoken but correct language, domain vocabulary used naturally, and once per text room for a deeper thought — one sentence that follows from the subject, without preaching.
Resistance to detectors does not rest on a list of banned words but on a list of banned patterns: paragraph symmetry, triads, "not only… but also", an intro with no content, a summary ending, walls of lists, an even sentence rhythm, no specifics.
What you getDraft and research, title options, shortening to a meta description.
Fact-checking, rewriting in the persona’s voice, cutting the patterns that give a generator away.
Proof: A ban on inventing facts, data, quotes, opinions and testimonials is part of the service. The persona is a style, not a licence for fiction.
The panel only handles content, and the site is a separate application. There are no plugins breaking each other after an update, and no single process where a panel failure also takes the site down.
Yes — one of our sites runs nine: articles, authors, categories, tags, comments, pages, catalog entries and two levels of catalog categories.
We estimate in hours, after a phone call and a look at the project. Our latest projects took 15 to 45 hours.
