When your AI assistant already has a page as HTML (an artifact it made, a design, an exported page), it doesn't need to build the page widget by widget. It sends the whole page in one go (import_page), and ContentForge builds it.
What happens
- ContentForge opens the page the way a visitor would, at desktop, tablet and phone sizes. It reads the layout, fonts, colors, spacing, hover effects and animations.
- It turns each part into our own widgets, the same way a website import does: for example cards, a card carousel, a scrolling text strip, a form or an accordion. Arch and circle photos, color tints over photos and thin divider lines come across too.
- Every picture the page uses is added to the site's media library in one batch.
- The page is saved to the draft. Nothing goes live until you publish.
It's free: no Zeus and no credits are used. A page usually takes seconds.
Where it goes
A new site for a client, with this page as its home page. The design's header, footer, phone menu, fonts and colors become the site's.
Replace a page on an existing site (for example the home page). The site keeps its own header, footer and theme unless you ask for the design's.
A new page on an existing site.
"Make this artifact the home page of a new site for Salem Counseling."
"Replace Smith Roofing's home page with this design, and use its header and footer too."
Fonts and phones
- Headings and body text use the site theme's fonts, picked from the fonts the page uses most. So when you (or your assistant) change the heading font in the theme, the imported headings change too. Only text in a different font, like an accent word in a script font, keeps its own font.
- On tablets and phones, the page follows the design's own layout for that screen: rows that don't fit side by side go one under another, the extra space a row had on desktop isn't kept, and the header stays on one line when the design's phone header does.
Your words stay exactly as written
The words on the page are saved exactly as they are in the design, even where they go against a writing rule (a "Learn more" button, for example). Your assistant is told to ask you before changing any of them.
Pictures
Your assistant uploads the page's pictures first, straight from its own workspace, so it doesn't have to write them out (which would be slow and use up your usage). Then it imports the page, and ContentForge finds each picture by its file name. A tiny picture or two can also go inside the call.
If your assistant runs in its own sandbox (a chat app in the browser, for example), allow your ContentForge address once in that app so pictures can upload. See the setup article for your assistant (Settings → API & MCP lists them). If an upload is blocked, your assistant tells you this in one sentence.
A picture the page points to by web address, with no file sent, is copied into the site's media too. If one can't be copied (blocked, too large or gone), it keeps its address so nothing is blank, and the result names it with the reason.
If pictures the page needs are missing, the result says so first and names each one. The page is still built, and a missing picture never takes its color or gradient with it. Your assistant uploads the missing pictures, then imports the page again, and they're connected.
Afterwards
The result lists any part that couldn't be turned into one of our widgets, with the reason and the closest widget. That part is still on the page as plain boxes, text and pictures. Your assistant can touch it up with the page editing tools, or you can in the editor. To put the old page back, ask your assistant to undo the change.