We’re building Ever Works in public. Some things may be incomplete, missing, or broken while we continue improving the platform.We’re building Ever Works in public — expect a few rough edges.
Builder
Nothing in Ever Works starts from a blank page. Every Work is generated from a maintained template — a complete Next.js or Astro codebase — and filled with researched content, so what you build on is a working site rather than a mock-up. You describe what you want in plain language; the runtime clones the template into your Git and fills it.
One prompt box, one form, and a run you can watch. Nothing is created until you press Create.
Click + New, say what you are building — the audience, the tone, what it must include — and pick a kind: Website, Landing Page, Blog, Directory or Awesome Repo. Your prompt carries through to the create form, so you never type it twice.
The Template field lists your own templates first, then the public blueprints for that kind; the catalog’s production entries today are Directory, Directory (Minimal) and Marketing Site. Pick nothing and the Work resolves to your saved default, or else to the base its kind starts on.
In the same form, choose the Git provider and the deploy target — Ever Works managed hosting, Vercel or your own Kubernetes cluster — and open Advanced to choose the research pipeline and the AI provider.
The runtime clones the template into a website repository in your account, researches the topic, and writes items, categories, tags and collections into a data repository. Watch it on the Worker tab, and cancel mid-run if the brief was wrong.
Directory and awesome-repo Works start on classic; website, landing-page and blog Works start on web. The other two are opt-in, lighter siblings.
Search, faceted filters, item pages, categories and tags, comparison pages and community submissions, on the App Router with Tailwind. The largest base, and the default for directories.
The same directory shape with as little JavaScript as possible: static output, interactive islands only where needed, and SEO, filters, search, sitemap and RSS each a plugin you switch on.
A complete landing page — hero, features, how-it-works, pricing, FAQ and call to action — plus About, Pricing and Contact pages and a Markdown blog, with every string in one site config file.
The same marketing sections, built statically with zero JavaScript by default. Copy lives in one site config file, and the theme is a plain-CSS surface an agent can restyle for you.
A directory Work is not a page-builder export. The classic template arrives as a real application, and the runtime keeps its content accurate.
Debounced search with URL-synced state, category and tag filters, sort and pagination. Every item sits in exactly one category and at most one collection, kept as plain YAML in your data repository.
Head-to-head comparison pages that score two items dimension by dimension, with their sources — switched on per Work. Editorial collections such as “Editor’s Picks” are on from the first run.
Two verdicts per item: whether the link is reachable, and whether the page is genuinely the right, official source — with the reasoning stored and a replacement URL suggested.
Visitors submit an item by pasting a URL, with moderation queues, item review and a full admin panel behind real role-based access control.
Subscription plans and sponsor ads with an approval workflow, through Stripe, Lemon Squeezy, Polar or Solidgate, plus a promo-code system for discounts on your listing pages.
Sitemap, robots, OpenGraph and JSON-LD, RSS, Atom and JSON feeds, llms.txt, and 21 languages with hreflang tags — generated on the first build and on every one after it.
A template is where you start, not a cage. Fork a built-in base into your own GitHub account and make it your default, or register any GitHub repository as a template of your own. On the minimal, web and web-minimal bases you can describe the look you want and an agent restyles the theme for you. And because the result is ordinary code in your Git, nothing stops you editing it directly.