Figma opens design files to AI coding agents via MCP

Figma has added an MCP server that lets AI coding agents write to and edit Figma design files while staying aligned with team components and styles.

The server works through a “use_figma” tool and supports MCP clients, including Claude Code, Codex, and Copilot.

Figma also introduced “skills,” Markdown instructions that teams can create without plugins to guide how agents generate and revise assets in Figma.

Executive from Codex and Claude Code expect tighter link between agents and the canvas to speed iteration.

The MCP server will be a paid API, but it is free during a beta period.

🔗 Source: Figma

🧠 Food for thought

Implications, context, and why it matters.

The workflow supports a “roundtrip” between code and design

AI agents can write to Figma. Another capability captures live web user interfaces (UIs) and converts them into editable design layers, not static screenshots 1.
This enables a “roundtrip” workflow. Software developers can send rendered UI into Figma as editable frames, refine layouts visually, then pull design context back into the coding workflow through the Figma MCP server 2.
Access varies by seat and plan. The remote MCP server works across all seats and plans, while read tools face rate limits. Starter plan users, plus View or Collab seats on paid plans, can make up to six tool calls per month. Dev or Full seats on Professional, Organization, or Enterprise plans have per-minute limits that match Tier 1 Figma REST API limits 3.

This integration keeps design systems central, but doesn’t make them the only “source of truth”

Agents can write to the canvas and produce design-informed code using structured design context such as components, variables, and layout data. This tightens the link between design files and implementation work 4.
This keeps the design system, including components and variables, as a source of truth that agents can reuse when creating or updating assets in Figma 5. The GitHub team used the Figma MCP server with its Primer design system (GitHub’s internal set of reusable UI components and design rules) to automate token sync and keep design and code aligned 6.
This approach strengthens Figma against AI tools that generate only images. It relies on structured design data and design-system-aware workflows rather than flat visual outputs 7.