Figma is enhancing the interaction between artificial intelligence models and design data with its new server, the Dev Mode Model Context Protocol (MCP) server, now available in beta for users with Dev or Full seat Figma access.
The MCP server is designed to provide large language models (LLMs) and agentic coding tools, such as Copilot in VS Code and Claude Code, with more precise contextual information about designs created within Figma. This dedicated developer tool goes beyond visual interpretation of image files, allowing AI models to directly access granular design data, including specific details such as precise numerical values used in graphs or the exact color shades applied in a design.
Figma illustrates the concept by comparing it to replicating a painting. Just as a painter needs both the visual reference and an understanding of the techniques employed, AI models utilizing the MCP server can access both the visual design and the underlying structural and data information. This is expected to improve the accuracy and efficiency of AI coding tools in translating designs into functional applications that closely mirror the designer’s original intent. The company anticipates this process will help reduce LLM usage by minimizing the guesswork involved in design-to-code transformations.
Figma has indicated plans for several updates to the Dev Mode MCP server in the coming months. These updates are expected to include features such as remote server capabilities and more profound integrations with existing codebases.
The beta release of the MCP server follows recent developments from Figma. The prompt-to-code Figma Make platform, initially introduced in May, became fully available to all Full seat Figma users on Tuesday. Figma Make allows users to generate functional applications simply by providing descriptive prompts. Additionally, the Figma Sites Code Layers feature, designed to provide AI tools for converting designs into interactive website experiences, is scheduled for rollout on June 12th.
