Figma opens canvases to AI agents

Figma now lets artificial intelligence agents place and edit designs directly on the canvas, instead of only reading files for code generation. (figma.com) Figma announced the change on March 24, 2026, as an expansion of its Model Context Protocol server, or MCP server, which first launched in beta on June 4, 2025. The new `use_figma` tool can create and update assets inside Figma Design files using existing components, variables, and auto layout rules. (figma.com, figma.com) Model Context Protocol is a standard that lets an artificial intelligence tool connect to outside software and pull in structured context. In Figma’s case, that context includes components, variables, layout data, and other design details that describe how a product is built, not just how it looks. (figma.com, help.figma.com) Before this update, Figma’s MCP work was aimed mainly at giving coding tools design context so they could generate more accurate code. With write access, the same agent can now move in the other direction and build or revise frames, components, and screens inside Figma itself. (figma.com, help.figma.com) Figma says the agent does not have to invent a brand system from scratch. The company says agents can build with “real Figma primitives” and reuse the components and variables already stored in a team’s libraries, with the design system acting as the source of truth. (help.figma.com, figma.com) Figma also added “skills,” which are markdown instruction files that tell an agent which steps to follow and which conventions to obey. The company says those skills can encode team decisions about naming, sequencing, and workflow so the agent’s output follows local rules instead of generic defaults. (figma.com, help.figma.com) The feature is aimed at the same tools many developers already use around code. Figma’s MCP catalog lists read-and-write support for clients including Claude Code, Cursor, Visual Studio Code, Codex by OpenAI, Windsurf, Warp, Firebender, Factory, and Augment, while some others remain read-only. (figma.com) Access is still gated. Figma’s help documentation says the new write-to-canvas feature is available to Full and Dev seats on paid plans, with Dev seats limited to read-only access outside drafts, and the company says the capability is free during beta but is planned to become usage-based later. (help.figma.com, help.figma.com) Alongside the agent update, designers have also been circulating a free Figma Community file called “Launch UI,” a landing-page component and template kit by creator Mikołaj Dobrucki. The file is described as free, editable, and ready to use for landing pages, giving teams a starter canvas for testing agent-assisted page assembly. (figma.com) The combined pitch is straightforward: connect an agent to Figma, point it at a real design system, and let it assemble first drafts inside the same file where humans will review and refine them. Figma says that keeps code, canvas, and brand rules in one workflow instead of splitting them across separate tools. (figma.com, help.figma.com)