Figma: Build by design — new workflows for building with AI
At TechCrunch Disrupt and Config India on October 15, 2026, Figma CEO Dylan Field published a strategic post that frames the company’s latest releases around a single tension: teams now build in two distinct modes, and most tools handle only one of them well.
Field calls these modes “building to think” and “building to ship.” The first is exploratory — rough prototypes generated quickly to test an idea, often by one person without visibility into why certain decisions were made. The second is deliberate — careful, detail-oriented work aimed at a shippable result. The problem, he argues, is that AI has made the first mode very cheap, which has made the gap between exploration and production larger, not smaller.
Figma’s answer is a set of October launches designed to keep both modes in one shared environment:
Code Layers (open beta, all plans) lets teams build code-backed prototypes directly on the canvas. You can import a local folder, pull from GitHub, connect via MCP, or capture a live website through the Chrome extension. The resulting prototypes stay in Figma, visible to the whole team.
Specs is a Markdown-based editor that lives in Figma and holds shared context — product requirements, technical notes, diagrams, links to relevant designs. Through the Figma MCP server, coding agents can read these specs when generating code, giving them the same context a human designer would have.
Design system tools expand in two directions. Guidelines and code sources help agents follow established component patterns. A new system map shows how libraries depend on each other, and system checks flag missing descriptions or incomplete interactive states.
Payload (the open-source CMS whose team Figma acquired) is rolling into Figma Make, so teams can manage content and data alongside app and website building.
Figma agent is now generally available after its beta. Twenty new MCP partner integrations have been added, including JetBrains Junie, Meta Muse, and OpenCode.
For designers, the practical significance is that the exploration phase no longer needs to happen outside Figma. AI-generated prototypes can live on the same canvas as production designs, with context (from Specs) and design rules (from Guidelines) available to whatever coding agent picks up the work next.