Figma adds Specs and code tools to guide AI agents

Figma is giving design teams a more structured way to work with AI agents, bringing design-system guidance, code and Markdown-based context into the same canvas. The company announced the changes at its Config India event, with the product details also reported by TechCrunch.¹²

Figma Specs puts design context in one place

The headline addition is Figma Specs, a native Markdown-based space for documenting a company’s design language. A Spec can exist as its own file or alongside a Figma canvas, while also referencing canvas elements, FigJam diagrams, code layers and designs.¹ That gives teams a shared reference point for the rules and decisions an agent should follow, rather than leaving those instructions scattered across separate documents.

Figma is also expanding the context available to its design systems. Teams can add design libraries, code libraries and guidelines, while agents can use that centralised information when creating or editing components and code layers. Guidelines are moving into open beta, while code sources are starting in closed beta for paid plans.¹

Code layers and Payload integration

Code layers, first announced in private beta, are moving into open beta on all plans. Figma says users can turn designs into interactive code-backed prototypes, bring in code from a local folder, GitHub or MCP, and iterate on the result with their team on the canvas.¹ The move makes the canvas less of a hand-off point and more of a shared workspace for design and implementation.

The company is also integrating Payload, the open-source headless CMS it acquired last year, with Figma Make for dynamic websites and content management. Figma says Payload downloads have increased sixfold over the past year.²

Figma said its MCP ecosystem now supports more than 20 tools, including Open Code, Dia, Grok Bot and Sarvam. That wider compatibility matters because the company is positioning its canvas as a place where people and agents can share context, rather than simply as a visual editor.²

Our opinion

Figma’s most useful idea here is not another AI button; it is the attempt to make design intent legible. Specs, guidelines and code sources could give agents a better brief before they start changing a product. The open question is whether teams will maintain that context carefully enough for it to stay trustworthy. If they do, Figma’s canvas starts to look less like a presentation layer and more like a working map of how a product should be built.