Figma has recently launched a series of new updates aimed at enabling AI models to communicate directly with its application building tools and access designs remotely. The core of these updates is Figma's Model Context Protocol (MCP) server, which acts as a bridge allowing AI models to directly access the code behind prototypes and designs created in Figma tools.

image.png

According to Kris Rasmussen, Figma's technical lead, the MCP server has been expanded to support Figma's AI prompts to the application coding tool Figma Make. By accessing Figma Make files through an MCP client, AI models can view the underlying code, not just the rendered prototype or image. This means the MCP server can accurately show AI models and agents how applications are built in Figma Make, eliminating the need for guesswork based on visual designs.

Figma states that the Make MCP server experience is already supported by products such as Anthropic, Cursor, Windsurf, and VS Code, and future updates will allow third-party MCP servers to connect with Figma Make. The remote access functionality of the MCP server makes it easier for AI coding agents, browser-based models, and developers of integrated development environments to use the tool, as they previously needed to access Figma's desktop application locally.

In addition, Figma plans to introduce some new features, including a design snapshot feature that can convert Make snapshots into editable layers in Figma Design, a feature expected to launch this week. Meanwhile, Figma is testing an editing feature that allows users to manipulate designs directly on the design tool canvas using AI prompts, further enhancing the user's design experience.

Key Points:

🌟 Figma's MCP server enables AI models to directly access and understand the underlying code of application designs.  

🚀 The update to Figma Make will support multiple products, enhancing interoperability between AI and design tools.  

🖌️ New features will simplify the design process, including converting snapshots into editable layers and using AI prompts for design manipulation.