Figma ties AI to real systems
Figma introduced 'Make Kits' and 'Attachments' so AI-generated interfaces can respect actual component libraries, code and documentation instead of producing one-off mockups. That shifts the value to designers who can author and manage design systems — not just those who can make attractive pages — because generated UIs will now be judged by implementation fidelity as well as look. (aiposthub.com)
Most artificial intelligence interface tools have worked like a talented intern with no access badge: they can draw a nice screen, but they do not know your company’s button rules, naming conventions, or code. On April 2, 2026, Figma said its Make tool can now pull in that missing context with two new features called Make kits and Make attachments. (figma.com) A design system is the rulebook behind an app’s screens: which button sizes exist, which colors are allowed, how forms behave, and which components engineers already ship in code. Figma’s pitch is that Make kits let teams import those rules so generated prototypes start from real system parts instead of generic rectangles. (figma.com) Figma says a Make kit can include public npm packages, private code packages published inside a Figma organization, and Figma library styles, variables, and tokens. That means the model is not guessing what a “primary button” looks like; it can be pointed at the exact package and design tokens a team already uses. (figma.com) Attachments solve a different problem. They let a prompt travel with supporting files like product requirements, screenshots, spreadsheets, or research notes, so the generated prototype has actual source material instead of a one-line description. (figma.com) Figma’s own example is simple: ask for a checkout flow with a specific payment method, then attach the requirement doc and the right kit. The result is supposed to follow the company’s components, copy patterns, and constraints in one pass instead of needing a cleanup round after the “magic” demo. (figma.com) The company is also telling teams that these kits need written instructions, not just code. In Figma’s developer docs, it says authoring guidelines for a Make kit should read like the documentation you would hand a new engineer so they know how to use the system correctly. (figma.com) That changes who gets leverage from generative design tools. If the model can already produce a polished-looking landing page, the scarce skill moves upstream to the people who define components, tokens, usage rules, and exceptions well enough for a machine to follow them. (figma.com) It also pulls design closer to engineering. Figma says Make kits are rolling out to full-seat users on paid plans beginning March 26, 2026, and organizations can approve published kits for everyone, which turns the design system into shared infrastructure instead of a private file a few designers maintain by hand. (figma.com) The quiet part of this launch is that “looks right” is no longer enough. A generated screen now has to survive contact with the real product stack — the package names, the variables, the documented states, and the edge cases in the attached files — or it is just another mockup that dies before handoff. (figma.com) So the winning designer in this version of artificial intelligence is not the one who can prompt the prettiest frame in 10 seconds. It is the one who can build a system precise enough that a model, a teammate, and eventually production code all end up speaking the same language. (figma.com)
