Figma Blog: The Figma canvas is now open to agents

On March 24, 2026, Figma announced that its canvas is now open to AI agents. Where agents previously could only observe designs — reading content, describing layouts — they can now act directly on canvas elements.

The system works through a modular architecture called Skills. An agent’s capabilities within Figma are defined as discrete named functions it can execute. Teams can configure which agents have access to which Skills, allowing organizations to set boundaries on what automated tools can touch.

This matters for designers because it shifts agents from passive observers to active collaborators. An agent handling component renaming, style token enforcement, or repetitive layout tasks is meaningfully different from one that only describes what’s on screen. The distinction between read access and write access on a design file has significant workflow implications.

This announcement follows Figma’s earlier MCP integration and continues the platform’s direction toward workflows where agents participate in design processes rather than sit outside them. The practical impact depends on which integrations partners build and which Skills they define — but the underlying capability, giving external systems write access to the Figma canvas, is a structural change in how the tool can be used.