Figma adds Markdown-based Specs and Payload CMS integration to steer AI agents

At its first Config event in India, design platform Figma announced a set of features aimed at shaping how AI agents operate on its platform, according to TechCrunch. The announcements centre on Figma Specs, a Markdown-based editor for codifying a company's design language, alongside a public beta of the code layers feature and a tighter integration with the Payload headless CMS that Figma acquired last year.

Figma Specs turns design rules into agent-readable context

Figma Specs is a Markdown editor that lets teams write down their design language in plain text and treat that document as a shared reference for both people and agents. According to TechCrunch, a Spec can live as a standalone file inside Figma or sit alongside work on the canvas, and it is not limited to text: users can point to elements on the canvas and embed FigJam diagrams, code layers and existing designs.

Beyond the document itself, teams can attach further context such as design libraries, code libraries and guidelines, effectively building a centralised description of a design system. The stated goal is that agents, whether Figma's own or third-party ones, consult this system and keep their output consistent with a brand's rules during generation.

Code layers open up in public beta

Figma also moved its code layers feature from private beta, where it had sat since June, into public beta at the India event. Code layers attach working code to designs directly on the canvas, making prototypes more interactive, and users can ask the on-canvas agent to convert a design into code or modify a specific part of the code to iterate on a prototype.

The company had rolled out its on-canvas agent to a wider audience in the past week. TechCrunch reports that because AI output is inherently unpredictable, Figma is positioning these new definition tools as a way to rein it in and move toward more predictable results.

Figma's chief design officer Loredana Crisan framed the shift in terms of how much code now serves exploration rather than shipping, telling TechCrunch that "there are 17-times more lines of code being produced today than in the past" while releases have grown only 1.3 times. Much of that code, she argued, is "built to discover what the product is" — in effect, code used as a design medium.

More tools plug into Figma's MCP

Figma said it now supports more than 20 additional tools through its MCP (Model Context Protocol) offering, including Open Code, Dia, Grok Bot and the Indian AI startup Sarvam, according to TechCrunch. The expansion widens the pool of external agents and coding assistants that can read and act on Figma context.

Payload lands inside Figma Make

The Payload headless CMS, which Figma bought last year, is being integrated with Figma Make, the company's product for generating dynamic websites. The integration is intended to let teams generate sites and manage their content from the same environment. Figma also said Payload downloads have grown sixfold over the past 12 months.

Why India is the venue

The announcements were made at Figma's first Config event in India, a market the company describes as its largest by user count outside the United States. Figma opened its first Indian office late last year with sales, partnership, go-to-market and product support functions, and says its platform is used by 40% of India's top 100 companies listed on the Bombay Stock Exchange.

Why it matters

As AI agents take on a growing share of design and code generation, the bottleneck is shifting from producing output to constraining it. Figma Specs and the surrounding context tools are an attempt to make brand and design-system rules machine-readable, so agents produce work teams can actually use rather than discard. The Payload integration points in the same direction for content, keeping generated sites connected to structured data after the first render. For design teams, the announcements signal that governance, not generation, is becoming the surface where tools like Figma compete.