Published 24 July 2026 · Source dated 24 March 2026
Figma's AI agent now works natively on the canvas, generating layouts grounded in your real components and tokens rather than flat mockups. It's shipped, not speculative — but Penpot users won't get the same feature yet.
Figma's canvas agent can now work directly inside your files — exploring layout directions while staying grounded in your existing components and design tokens, instead of just generating flat images to rebuild by hand. It's a shift from AI-as-image-generator to AI-as-collaborator inside the actual tool. For junior designers on Figma, this means AI suggestions start appearing as real, editable layers you can inspect and tweak, not just inspiration to eyeball. Penpot doesn't have an equivalent yet, so if you're using both tools day-to-day, know this is a Figma-specific gap for now — worth factoring in if you're choosing which to build a portfolio piece in.
Coach
Open a Figma file this week, try the design agent on a small component set, and note one thing it got wrong — that's solid interview talking material.
Figma's MCP catalog turns AI agents into real clients
Figma has published a live catalog of AI coding tools (Claude, Cursor, VS Code, Grok, Google Antigravity and more) that can connect to Figma via its MCP server, with several getting read-and-write access. It's the clearest sign yet that connecting an AI agent to your Figma file is becoming normal practice, not a hack.
figma.com ·
Adoption
Figma's MCP Server Ships Real Team Workflows
Figma just shared four concrete internal use cases for its Dev Mode MCP server, moving it past demo territory. It's now stitching together decks, Code Connect, and Figma Make output with actual codebases.
figma.com ·
Adoption
Figma Adds 3D Transforms & MCP-Ready Motion
Figma's July 2026 release notes confirm 3D transforms for frames, vectors and text, exporting straight to CSS. Motion frames are now MCP-compatible, so coding agents get full animation context instead of a flat screenshot.