XAIO 1.2: prototypes, website capture & publishing without guesswork

Version 1.1 was about building with the best models and rolling AI development out to a whole organization. Version 1.2 is about the two ends of that work: exploring an idea before you commit to it, and shipping without hitting a wall you cannot see.

Sketch first, then build

While you are still exploring, nothing needs to be rebuilt for real. The assistant now sketches self-contained UI prototypes into a new Prototypes tab. Compare them, discard them, and when one is right, “Build this” hands the sketch to the chat as a request you can still edit before sending.

Bring an existing website in

Give the assistant a URL and it captures the site through the XAIO Browser integration — pages and assets land in your project with a live progress card in the chat. It is the starting point we kept being asked for: a relaunch or a migration that begins from what you already have, rather than from an empty project.

Publishing that names the blocker

A first publish used to dead-end at a greyed-out button with no explanation — most often because the suggested address was already taken. Now the main action reads “Finish setup” above a list of exactly what is still needed, and one click opens the right settings with the required fields marked. Saving hands you straight back to the publish dialog.

Underneath that, every project owns a deployment target from the moment it is created — inactive, with no address and no cost — and a target cannot reach the publisher half-configured any more: name, hosting and SEO title are required, checked both on save and on publish.

Documentation that stays with the code

Every project now has a docs/ folder next to frontend/ and backend/, and a Documentation tab that shows it. Markdown renders as a document with a toggle to the source. Because the folder lives in the project’s own repository, every document is versioned with the code it describes and travels with a clone, an export or a publish — and the assistant reads and writes it as part of ordinary work.

For organizations

• A monthly AI-task budget per member, so one person cannot quietly consume the shared pool.
• Platform Builder trials: a two-step request that works logged out, and a granted trial provisions a ready-to-use organization with its own AI-task pool — no credit card.
• The model policy is now enforced by the services themselves, not only by the picker in the workspace.

And a bigger free start

The Free plan now starts with 250 AI Tasks for 30 days instead of 50 a month, with no daily limit inside that window — enough to actually finish a first project. The month begins with your first build, not with signing up. Everyone already on a paid plan received 250 AI Tasks as a one-off thank-you; they sit in the top-up balance and never expire.

Everything else

The full release notes live on the Version 1.2 page.

• Kimi K3 joins the model picker, and the chat names the model that actually answered.
• Attach any file type to the chat — documents, archives, data files — with a preview card per file.
• When a turn fails, the message names the cause instead of “stopped unexpectedly”.
• Review pre-release findings, fix the ones you pick with AI, then resume the held publish.
• Project rules on mobile, read-only for collaborators; environment values masked with a reveal toggle.