Vibe code your own Webflow
Vibe code your own Webflow as section pages, one collection, and a publish. Basic is $15 a month billed yearly. The visual canvas stays on the long row.
You can vibe code your own Webflow as section pages, one collection, and a publish button. Webflow is that product. The catalog line is “A few pages, a list of posts, and a publish button.”
The card lists $15 to $39 a month. Webflow’s pricing is the source for that range. Basic is $15 a month billed yearly, through Premium at $39 a month billed monthly. The card leaves out Ecommerce, Team, Workspace, and Enterprise quotes. The first estimate is 12 hours. The verdict is “Pages and a collection”.
Last time I wrote about vibe coding your own OptinMonster. That was a regular post. This one is a teardown.
The public sources are Webflow’s scaling post (September 10, 2024), the CMS scale post (October 4, 2022), and Designer APIs, part 2 (September 25, 2024). Where those pages stop, I call it a guess.
Vibe code your own Webflow from one collection
The Webflow page splits the work into three rows.
The first row is 12 hours. Section pages, one collection, and publish. You edit fixed fields. You publish. The public site shows that release.
The second row is 4 days. Drafts you can preview, redirects, a second collection, and a custom domain.
The third row is 1 month and up. A visual canvas, interactions, ecommerce, and a marketplace. The card calls that row the part to skip.
If you tell an agent to “clone Webflow”, it aims at the third row. Copy the prompt, then pin it to the first one.
Store SitePage, Section, Collection, CollectionItem, and SiteRelease. The card names all five. A page is a slug, a title, a description, and an order of sections. A section is a type you already defined, plus the fields that type allows. A collection is one list, here notes, with title, slug, body, and a publish state. A release is the published snapshot the public site is allowed to read.
The audience is one founder keeping a marketing site and a short blog. The same v1 also stands in for Typedream and Super.so. The card lists both and does not give them prices.
A freeform page is Framer at $10 to $30 a month, the monthly equivalents of Basic through Pro billed annually, on Framer’s pricing. One stacked page is Carrd at $2 to $4 a month, in Build a Carrd alternative. A landing page and a form is Unbounce at $29 to $249 a month, on Unbounce’s pricing. This v1 is a home page plus a notes list.
The guardrail is the rule you paste into the prompt. Serve only the published release. Accept structured section fields, not arbitrary code. Keep drafts off guessed URLs.
How does Webflow publish a site
Leah runs northlamp.com, a studio site. On November 8, 2026 she has one page, Home, and one collection, Notes.
Home is three sections. The hero says “Lamp hours” and “The studio is open on Tuesdays.” The feature block is two sentences about the November class. The footer is the address and one email, [email protected]. Notes has two items. “Lamp hours” is published. “A November note” is still a draft.
She hits publish. The app writes a release that holds Home and the published note. The public site renders that release. The draft note is not in it. A guessed URL for the draft returns nothing.
That is the 12-hour loop. The pages she can build are the section types you listed in the prompt. She cannot drop an arbitrary box on the canvas, and she cannot write a style rule by dragging.
Webflow’s Designer is the thing that loop leaves out. The Designer APIs post says V2 talks to the Designer on every call. The app and the Designer pass messages with postMessage. Each call updates an internal Flux store, which they describe as their version of Redux, and that store is what gets persisted. Elements, pages, and styles are the records. V1 held the same edits in memory until something called save(). V2 can also create elements from Element Presets, because the call now runs the Designer’s own logic instead of a batch you flush later.
The scaling post says what publish does with that design. Webflow uses AWS Lambda to convert and render static pages and CMS content into HTML, CSS, and JavaScript, ready to view globally within seconds. In an hour they see several hundred new sites on the canvas, more than 142,000 style applications on an HTML element, and more than 15,000 publishes, about one every four seconds. The same hour runs more than five million Lambda functions, with a failure rate under 0.001%. About 15,000 images are uploaded in that hour. The API side is more than 10 million requests an hour, stored in a document database on AWS S3: over two petabytes and nine billion objects, at many tens of thousands of operations a second.
The CMS post, from October 2022, is about how many items that render can hold. Performance had kept customers at 10,000 CMS items or fewer. They describe a new storage layout for Collection content, built so Enterprise sites can go past 10,000, 20,000, and 100,000 items, plus more collections. They also say Designer load times improved earlier that year, and publish got up to 10 times faster on the largest projects. The post says the new layout stores and retrieves Collection content more efficiently, and that the structures expand as the customer base grows. It does not name the tables or the indexes. The shape of that store is a guess.
Interactions, ecommerce, and a template marketplace are on the card’s month row. None of these three posts describe them, so how those features are built stays a guess. The scaling post does say image uploads are optimized. It does not describe a storefront.
Your browser, on the edit screen, talks to one Astro app. Node.js plus Astro is the default. Use TanStack Start if React is your weekday language. The capability line on the card is Astro, HTMX, and D1. Pages, sections, the collection, and the release live in D1. HTMX, or a plain form, is enough to edit a hero and a note. The decision to publish stays on the server. A static hosting hook can ship the release. That hook is the integration the card names, and it is still one snapshot, not a canvas.
Screens of a Webflow alternative
v1 needs the page editor, the collection, and the public page. A preview link and a second collection are the 4-day row.
The page editor
Leah sees Home as three blocks she can edit, not a canvas. Publish writes the release. The draft note is not a block on this screen.
The notes collection
The list is title, slug, body, and a state. “Lamp hours” is published. “A November note” stays a draft until she changes that state and publishes again.
The public page
northlamp.com shows the hero, the feature, the footer, and the published note. The draft is absent. There is no editor chrome on this URL.
How a section becomes the live page
Leah edits the hero, then publishes. The server checks that every section is one of the allowed types and that the fields match. Home and the published note are copied into a release. The public route reads the latest release and nothing else.
A second publish replaces what the public route reads. The previous release stays in the table so a later rollback has a row to point at. The card’s feature list includes that rollback. Store the row on this pass. The button that points the public site back at it can sit with the 4-day draft screen.
Redirects, a preview hostname, and a projects collection are the 4-day row. A freeform canvas, click interactions, a cart, and a marketplace are the month row.
The hard half of a Webflow alternative
Twelve hours is a page of known sections, one notes collection, and a release the public route will serve. Leah opens northlamp.com and sees Lamp hours. She opens the draft slug and sees an empty response.
The hard half is the canvas and the fleet behind publish. The Designer API post says a call updates Flux, and that store is what gets persisted. The scaling post counts more than 142,000 style applications on an HTML element in an hour. It does not say each of those is its own database round trip. Publish is a render. Lambda turns static pages and CMS items into HTML, CSS, and JavaScript, more than five million runs an hour, beside 10 million API requests and a document store of nine billion objects.
Two notes fit in D1. A library past 10,000 items is the storage problem their 2022 post says they rebuilt, without publishing the schema. Interactions, a cart, and a marketplace stay off these pages. The card’s non-goal leaves them out, and Framer already covers a published page from sections.
Paste the field rule from How to write a coding agent prompt, and keep the split in How to scope a SaaS clone.
Start with the first row
Go to Build your own Webflow. Copy the Astro prompt, or TanStack Start if that is what you write.
Tell the agent the first estimate only. Section pages, one collection, and publish. Twelve hours.
Publish Home with the three sections, publish “Lamp hours”, leave “A November note” as a draft, and confirm the live page shows one note. If you need a canvas, a cart, or a marketplace, keep the plan you already pay for that part.