Figma debuts design agent

Figma on May 20 rolled out a beta version of its design agent, an AI assistant that works inside Figma Design rather than as a separate chatbot or plugin. The company said the agent can generate and remix designs, automate repetitive work and use existing design systems while operating on the canvas where teams already collaborate. Figma published a product video the same day featuring product designer Tammy Taabassum and product manager Rodrigo Davies, and paired it with release notes describing the tool as “purpose-built” for design. The launch extends Figma’s broader push to make AI agents operate on structured design objects instead of only producing text or flat images. Figma’s documentation says agents can read components, variables, layout data and other design details from files, and with new write capabilities can also create and update designs directly. (youtube.com) ### What exactly did Figma release on May 20? Figma’s May 20 release notes said “The Figma agent is here” and described the product as embedded “where design already happens.” The company said the agent is rolling out gradually in beta over the coming weeks and will be available to Full seat users on Professional, Organization and Enterprise plans, while Collab and Dev seats can use it in drafts. Starter, Education and Government plans are excluded from the beta, according to the release notes. (help.figma.com) TechCrunch reported on May 20 that the agent first launches in Figma Design and that Figma plans to bring it to other products later. The publication said users can direct the agent with natural-language prompts to generate new designs, edit existing work and automate tasks such as creating multiple iterations. (figma.com) ### How is this different from an AI sidebar that just suggests copy? Figma’s own product materials describe the agent as working with native design structures. Its help documentation says agents can extract components, variables and layout data from files, then build with “real Figma primitives” and reuse what already exists in libraries instead of producing only visual mockups. (techcrunch.com) The company’s March 24 blog post laid out the same architecture in earlier form, saying AI agents could design directly on the Figma canvas and be guided by “skills” that carry a team’s decisions and intent. Matt Colyer, a product director at Figma, wrote that agents had previously lacked that design context, which is why many AI-generated designs felt “unfamiliar and generic.” (help.figma.com) ### What are “skills,” and why do they matter here? Figma’s help center says skills are pre-built instructions that teach an AI agent how to handle common tasks reliably, including creating a new file, building a screen from a design system or generating code from a design. The company lists skills such as `figma-use`, `figma-create-new-file`, `figma-generate-design` and `figma-code-connect` for different workflows. (figma.com) The same documentation says `figma-use` is the foundational skill for writing content to a Figma canvas, including frames, components, variables and layouts. Figma says skills can be bundled through supported clients such as Claude Code or Cursor, or downloaded and installed manually. ### How does this connect design work to code? (help.figma.com) Figma’s developer documentation says its MCP server brings Figma into an agent workflow by giving models access to design information and letting them write native Figma content back to the canvas. The company says the server can also generate code from selected frames, extract design context into an IDE and retrieve resources from Make files. (help.figma.com) In its March 24 blog post, Figma said Claude Code, Codex and other MCP clients could generate and modify design assets linked to a design system through the `use_figma` tool. Ed Bayes, design lead at Codex, said in the post that Codex can use design context in Figma to help OpenAI teams build products more efficiently. ### Who gets access first, and what happens next? (developers.figma.com) Figma said the beta rollout will happen over the coming weeks and that the feature will not consume credits during the beta period. The company said AI credits will apply at general availability and directed users to a waitlist for early access requests. The product video published on May 20 remains the main public walkthrough of the new agent, and Figma’s release notes and MCP documentation are the company’s current reference points for eligibility, supported workflows and future pricing. (figma.com) (youtube.com) (figma.com)