Figma: how the FigCade was built using Make, Weave, and MCP
What the article is about
Figma’s blog describes how employees used three of their AI tools — Figma Make, Figma Weave, and the Figma MCP server — to build six browser mini-games for an internal April Fun Day celebration. The post is structured as a behind-the-scenes walkthrough, with specific team members describing what they built, which tools they used, and where the workflow sped up or required them to step in manually.
Context
The project is useful as a case study precisely because the Figma team were using their own tools under real production constraints, not a controlled demo. They had a fixed deadline, a small team per game, and the goal of shipping something that could actually be played. Each of the three AI tools served a distinct role in the workflow.
Key takeaway
The case study demonstrates a clear division of labor across the three tools. Figma Make handled rapid prototyping: one designer used it on a Sunday morning to validate the game concept for a quiz game called 2Fast2Figma, moving from idea to working prototype in an afternoon. Make is particularly effective at this stage because it generates interactive output directly rather than static screens, making it easier to test whether a concept feels right before investing further.
Figma Weave covered visual asset creation. Designer Lesley Moon used it to generate felt-textured cursor graphics in a couple of hours — work that would otherwise have required a specialist or a multi-day production process. Motion designer Fifi Law used Weave to generate storyboard elements for the April Fun Day trailer and shipped the full video in a single day.
The Figma MCP server connected the design work to code. Engineer Steven Noto used it alongside Claude and GitHub Copilot: by pasting links to specific design components from the MCP server directly into his AI coding environment, the agent could read the design specifications and generate code that matched those components exactly. Without the MCP connection, the engineer would have had to manually translate design intent into code requirements for every component.
What the case study does not show is how the team handled design decisions that required visual judgment — color corrections, spacing refinements, transitions — which appear to have remained human-led throughout.
Who should read this
Designers and developers who want to understand how Figma’s AI features work together in a real project rather than in a product demo. The post is also relevant for teams evaluating whether Make, Weave, or the MCP server would benefit their own workflows, since it shows which type of work each tool accelerates.