Datasets give a client's site data to read, and boards keep it current

Datasets, for the data a client's site reads

Boards have been doing double duty as databases. A CMS board for the blog, another for release notes, a third for whatever a client's website pulls in. It works right up until someone tidies a column name and a live page goes blank.

Datasets is a new area in the hub for data that sits still. Many records of the same shape go in a collection: blog posts, testimonials, a pricing table. One set of settings goes in a document, like the booking rules behind a contact form. You define the fields once, and every value is checked against them before it's stored.

Records draft and publish separately. Edit something already live and the client's site keeps serving the old version until you publish, so next week's copy can sit half-finished without anyone seeing it.

Each collection comes with its own delivery link. Hand it to whoever builds the site and they read the published records straight from it, no key, no trip through your hub. Rotate the link and that reader is cut off. The dataset also describes its own fields, so a developer can see the shape of what they're reading without waiting on you for a spec.

Renaming is safe. Change a field label, rename a select option, and whatever is already reading that field carries on, because the words on screen aren't what the data is filed under. Old versions of the shape are kept, so you can see what changed and when.

Everything here works over the API too, and through an AI agent connected to the org.

A card can keep a dataset current

Boards stay where work gets planned and reviewed. The dataset is what the outside world reads. Automations now connect the two.

There are fifteen new blocks. Five start a workflow when a record changes, and the rest read and write records and documents from inside one.

The interesting one is the sync block. A card reaches its publish date, moves to Published, and its properties land in the dataset behind the client's site. Run the same workflow over that card again and it updates the record it wrote last time instead of leaving a second copy, which is what makes it safe to schedule or to point at a backlog.

Fields pair up by name on their own. Where a board property and a dataset field genuinely disagree, you set the pairing. The pickers only offer datasets and fields you actually have, so a workflow can't be saved pointing at something that isn't there.

Automated posts to X can carry images

Publish to X was text only. It now takes up to four images from the card: an image property, a Media Center asset, or a Studio frame, all in the same field. Change a Studio design after it was last posted and the fresh version is rendered before the post goes out.

Video isn't supported on this route, and the workflow says so rather than posting without it. Nothing to reconnect, and posts already going out as text are unaffected.

Also in this release

Avatars load quicker across the hub: the account menu, organization marks, members, and customers.