Figma introduces new features to define AI agent workflows

Design platform Figma has rolled out several new features and capabilities aimed at defining how AI agents operate on its platform, steering them toward more predictable outcomes given the inherently non-deterministic nature of artificial intelligence. Among the key rollouts is a new feature called Figma Specs, which functions as a markdown editor allowing users to define their design language in a straightforward manner.

This documentation serves as a centralized reference for teams and agents alike during the design process in Figma. Specs can operate as a standalone file or exist directly within the canvas. In addition to text editing, users can reference other Figma elements on the canvas, embed FigJam diagrams, code layers, and designs, while also managing centralized context through design libraries, code libraries, and guidelines that establish rules for agents to follow.

Stock photo for illustration only, not from the actual event

Over the past week, the company has made its canvas-based agent available to a broader user base. Alongside this expansion, Figma is launching its code layers feature into public beta following a private beta phase that began in June. This feature introduces coding directly to designs on the canvas to build more interactive prototypes, enabling users to prompt the agent to convert a design into code or modify specific sections of the code to iterate rapidly on prototypes.