Figma Just Made AI a Real Design Partner
AI-generated UIs have always had the same problem. The layout looks right. The interactions feel smart. But the components are wrong, the data is fake, and the edge cases are missing.That is not a prototype. That is a starting point that takes hours to clean up before it is useful to anyone.
Figma just fixed that.
Make Kits: Teaching AI Your Actual Design System
Make kits let design system authors package their components, styles, and usage guidelines into a reusable context layer that Figma Make reads before generating anything.The result is immediate. AI output starts with production-aligned components instead of generic placeholders. Engineers recognize the patterns. Reviews move faster. The question shifts from "Is this custom?" to "Does this design work?"
Teams can pull in code components via npm or sync directly from existing Figma libraries. Forms, dashboards, onboarding flows. Anything that lives across shared surfaces now generates consistently, without designers manually correcting divergences after the fact.
Less cleanup. More building.
Make Attachments: Grounding Prototypes in Real Context
Design systems solve one half of the problem. Real project complexity solves the other.Make attachments let designers feed source materials directly into Figma Make. PDFs, markdown files, CSVs, JSON datasets, brand guidelines, legal copy, screenshots. Instead of describing a complex onboarding flow in a long prompt, you attach the actual spec and let the model work from it.
This matters most on the flows that fail in production. Validation states. Legal requirements. Real user data that reveals layout constraints you never considered in the abstract. AI used to simplify these because it had no other choice. Now it has the context to get them right from the first draft.
Prototypes stop being idealized mockups and start being working drafts.
The MCP Layer: Connecting Design and Code
Figma also launched an MCP server that brings design context directly into the developer workflow.The same components used in Make feed into code generation. That means engineers are not reinterpreting design decisions or rebuilding structures from scratch. They are reading the intent of the design and generating aligned code from it.
Less back and forth. Less "what did you mean by this?" in the review.
What This Actually Means
The gap between design and engineering has always been a context problem. Designers know things engineers have to guess. Engineers make calls designers have to undo.Make kits and Make attachments attack that gap at the source. AI is not generating a starting point you adjust toward your product. It is generating from inside your product.
That changes the speed of the design loop. It also raises the floor on what AI-assisted design can produce without human intervention.
The teams that build structured design systems and feed them into AI workflows are not just designing faster. They are compressing the distance between idea and production in a way that compounds over time.
If you are still generating from scratch and cleaning up after AI, this is worth paying attention to.
