Figma: AI Agents Can Now Write Directly to the Canvas
FigmaFigma launched an open beta enabling AI agents to write directly to the Figma canvas via a new use_figma MCP tool — the first time agents can create and modify native design assets (frames, components, variables, auto-layout) using a team's own design system as the source of truth. Alongside the write capability, Figma introduced Skills, markdown files that encode a team's design conventions and teach agents which components to use — with nine community-built skills shipping at launch. The feature supports nine MCP clients including Claude Code, Cursor, Codex, and GitHub Copilot, and is free during the beta period.
Sources & Mentions
5 external resources covering this update
Figma Just Opened the Canvas to AI Agents – Here's What It Means for Designers
Muzli
Figma Just Opened the Canvas to AI Agents — Here's What It Means for Designers
Medium
Figma AI Agents Now Write Directly to the Design Canvas
Abduzeedo
figma/mcp-server-guide: A guide on how to use the Figma MCP server
GitHub
Figma Release Notes – March 2026
Releasebot
AI Agents Come to the Figma Canvas
On March 24, 2026, Figma opened its canvas to AI agents in open beta — a milestone that shifts the Figma MCP server from a read-only design context provider to a full read-write design collaborator. Through a new use_figma MCP tool, coding agents can now create, edit, delete, and inspect any object in a Figma file: pages, frames, components, variants, variables, styles, text, and images. Critically, agents work inside a team's design system — not around it.
The core problem Figma is addressing is one familiar to anyone who has experimented with AI-assisted UI generation: AI-generated interfaces have historically felt generic because agents lacked access to the tokens, component libraries, spacing systems, and brand conventions that teams spend months building. The use_figma tool changes this. Agents now reach into the actual component library, apply real variables, and respect auto-layout rules — producing output that is structurally consistent with the rest of the product, not a pixel-approximation of it.
The use_figma Tool
The use_figma tool is the centerpiece of the beta. Unlike the existing generate_figma_design tool — which translates live HTML from a browser into editable Figma layers — use_figma is a general-purpose write interface. It enables agents to:
- Create and modify frames, components, and component variants
- Apply variables (color tokens, spacing, typography) from existing libraries
- Configure auto-layout and design constraints
- Screenshot output and iterate on visual mismatches
Skills: Teaching Agents Your Team's Conventions
The most distinctive part of the launch is Skills — markdown files that define how an agent should behave on the Figma canvas. Teams write Skills once; agents apply that knowledge consistently across every session. Writing Skills requires no plugin development or programming.
Nine community-built Skills launched on day one:
/figma-generate-library— Creates components from a codebase/figma-generate-design— Creates designs using existing components/apply-design-system— Connects designs to system components (Edenspiekermann)/create-voice— Generates screen reader accessibility specs (Uber)/sync-figma-token— Syncs tokens and detects drift (Firebender)/edit-figma-design— Orchestrates multi-step design workflows (Warp)/multi-agent— Runs parallel agent workflows (Augment Code)
Supported MCP Clients
The use_figma tool works with nine verified MCP clients at launch: Augment, Claude Code, Codex, Copilot CLI, GitHub Copilot in VS Code, Cursor, Factory, Firebender, and Warp.
Availability and Pricing
The use_figma open beta is available at no cost. Figma has indicated the feature will transition to usage-based pricing once the beta period concludes.