HomeAIAI AgentsFigma Opens Code Layers and Specs Betas and Adds Payload to Make

Figma Opens Code Layers and Specs Betas and Adds Payload to Make

Figma used its first Config conference in India on 15 October to push working code, written instructions for AI agents and a content backend further into its design canvas.

The headline change is code layers, which let designers place real, running code on the canvas next to their designs and turn them into interactive prototypes. Figma announced the feature in private beta in June. Chief executive Dylan Field said in a company blog post that it is now rolling out in open beta on all plans.

Teams can start a code layer by asking Figma’s agent to convert a design, cloning a connected GitHub repository, capturing a live web page with Figma’s Chrome extension, or starting from an empty layer, according to Figma’s help centre.

Specs gives agents a brief to follow

The second launch is Figma Specs, a Markdown editor for writing down a team’s design rules, product requirements and technical plans. A Spec can sit on the canvas beside the designs it describes and can reference canvas elements, embed FigJam diagrams and code layers, and be read or written by AI tools through Figma’s MCP server, the connector that lets outside agents work with Figma files. Specs is in open beta on web, desktop and mobile for paid plans.

Figma also widened what a design system can hold. Guidelines, written instructions attached to published libraries that are passed to the agent automatically, are in open beta on paid plans. Code sources, which link a library to a GitHub repository or npm package, are in closed beta. A single place to manage whole design systems, with a “system map” of dependencies, is in preview.

Loredana Crisan, Figma’s chief design officer, told TechCrunch why the company is leaning on code: “There are 17-times more lines of code being produced today than in the past. But when you’re looking at releases, only 1.3-times more releases.” Much of that code, she said, is built “to discover what the product is”.

Payload becomes Make’s backend

Figma acquired Payload, an open-source content management framework, last year. Payload is now built into Figma Make, the company’s prompt-to-app tool, giving projects a typed backend with a database, file storage, login, access control, an admin panel and APIs. Figma says Payload downloads have grown sixfold in 12 months. Payload is now Make’s native backend; Supabase remains available as a third-party connection, and existing Supabase-backed apps keep working.

The MCP server gained 20 new partner integrations, including OpenCode, the Dia browser, JetBrains’ Junie and Meta’s Muse, and Indian AI start-up Sarvam. Figma’s agent, which works directly on the canvas, became generally available on 6 October.

Figma said Config India drew more than 30,000 registrations, and that India is its second-largest market by daily active users. It says it works with more than 40% of the BSE 100, the Bombay Stock Exchange’s index of large listed companies.

4 November is when the beta starts to cost

The fine print matters as much as the features. Code layers run on Figma’s AI credits. Creating and editing inside them is free on paid plans only until 4 November 2026; after that every code-layer task draws credits like other agent features. Specs uses credits when the agent is involved, guidelines can make a prompt more expensive because they add context, and Figma says adding a Payload backend “uses more credits than building without it”. The agent itself has used credits since 6 October.

So the open beta is also a pricing test. Teams have until 4 November to learn what a code-backed prototype costs to iterate on, and whether that beats sending a designer’s work to an engineer or a separate coding tool such as Cursor’s agents. Only Enterprise customers can switch code layers off, and only by disabling all of Figma’s AI features, so most organisations will meet the feature whether or not they budget for it.

Specs is a bid to own the agent’s instructions

Markdown files are now a common way to give coding agents instructions; Anthropic’s Agent Skills format, which it has handed to an industry foundation, is built on them. Figma’s argument is that those files are scattered and cut off from the work they describe. Specs puts them on the canvas, linked to the designs, and makes them readable by any agent connected through MCP.

If that works, Figma becomes the place where a product’s rules are written down, not just where its screens are drawn, and every outside coding agent has a reason to connect to it. Design rivals are adding their own AI, as Canva did with its in-house Craft image model. Figma is betting that controlling the brief is worth more than generating the picture, and the 20 new MCP partners are the first sign of how many tools agree.

For tools that turn a screenshot rather than a Figma file into code, see Screenshot to UI Code: The Tools That Actually Do It.

Outside Figma, a separate set of tools starts from a picture rather than a design file. Open-source projects such as screenshot-to-code and hosted builders such as Vercel’s v0 accept a screenshot and return front-end code. TSN checked which screenshot-to-UI tools actually do the job.

Sources

Related stories

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

More in this category

Latest stories

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.