Figma Opens Beta for AI Agents That Design Directly on Canvas.

Figma has opened beta access to a new tool that lets AI agents design directly on its canvas, marking a shift toward agent-driven workflows in the design platform.

The company announced on LinkedIn that the use_figma MCP tool is now available in open beta, along with a set of skills designed to teach AI agents how to work inside the Figma editor. The launch follows recent AI image tools rolled out across FigJam, Buzz, and Slides.

The use_figma MCP tool is built on the Model Context Protocol, an emerging standard for connecting AI systems to external tools and data sources. By implementing MCP, Figma enables AI agents from platforms like Claude to perform design tasks directly on the canvas rather than just generating code or mockups.

The approach differs from AI-powered design assistants that suggest edits or automate repetitive tasks. Instead, agents equipped with the use_figma tool can execute design commands, manipulate objects, and build layouts based on natural language instructions or predefined workflows.

Figma is also providing skills, reusable instruction sets that teach agents how to accomplish specific design tasks within the platform. These skills act as training modules, giving agents knowledge of Figma's interface, components, and best practices. The combination of the MCP tool and skills aims to reduce the learning curve for agents and improve the reliability of their output.

The open beta is live today, though Figma has not disclosed detailed eligibility requirements or access limits. Developers and teams interested in testing agent-driven design workflows can begin experimenting with the tool immediately.

The launch positions Figma alongside other platforms exploring agent integration. Tools like Cursor, Windsurf, and Lovable have introduced agent-based coding and prototyping features, while Figma's implementation focuses specifically on visual design manipulation.

Whether the use_figma MCP tool will eventually support integration with no-code builders like Webflow or Framer, or expand to other parts of the Figma ecosystem such as prototyping and developer handoff, remains to be seen. For now, the beta centers on core canvas editing.

The move comes as Figma continues expanding its AI capabilities. Earlier updates introduced Make Kits to connect design systems with real project work, and support for 30+ file types in Figma Make. The use_figma tool extends that trajectory by opening the platform to programmatic control via AI agents.

Teams working with AI-driven design pipelines or exploring automation in their creative workflows now have a new option to test. Whether agent-based design becomes a production-ready workflow or remains an experimental capability will depend on how reliably the tool performs in real-world use cases.