Figma announced a set of tools meant to give AI agents clear rules to follow. The centerpiece is Figma Specs, a Markdown-based edit...
Figma announced a set of tools meant to give AI agents clear rules to follow. The centerpiece is Figma Specs, a Markdown-based editor where teams write down their design language. Alongside it, Figma opened its code layers feature to a public beta, added support for more than 20 new tools through its MCP, and is connecting the Payload headless content management system to Figma Make. The common thread is control. As more design work passes through AI, Figma wants a company's brand rules to live in one place that both people and agents can read.
What Figma Specs is
Figma Specs is a Markdown editor built into Figma. Markdown is a lightweight text format, so writing a spec feels like writing a plain note and not configuring software. Teams use it to define their design language in simple terms: how a brand looks, what its components are for, and what rules apply when something new gets made.
The document can serve as a reference for both teams and agents working in Figma. A spec can exist as a standalone file, or it can sit inside the canvas next to the designs it describes. It is more than a text editor, because a spec can point at other things in Figma. Users can reference elements on the canvas and embed FigJam diagrams, code layers, and designs. A written rule can therefore sit beside the example that shows it.
One place for design context
Specs also pulls in the material that usually lives in separate places. Teams can bring in design libraries, code libraries, and guidelines, and manage all of that context for the design system in one centralized spot. Agents can read from that same system to follow the rules.
This matters because design systems tend to be scattered. A color rule might sit in a style guide, a spacing rule in a component library, and an accessibility note in someone's head. A person can often piece that together from experience. An AI agent cannot, unless the rules are written where it can find them. Figma's pitch is that a central, readable definition gives its own agents and third-party agents a shared source of truth.
Steering a tool that is not predictable
Figma made its agent on canvas available to a wider group of users in the past week. The company is open about the catch: AI is non-deterministic, which means the same request can produce different results on different tries. For a brand, that is a real problem. A logo placed slightly wrong or a button in the wrong shade is the kind of drift a design system exists to prevent.
Figma's answer is to provide tools that steer the agent toward more predictable outcomes. Specs is the clearest example. It does not make the agent deterministic, but it gives the agent written rules to work from, which should narrow the range of results. How well that works will depend on how clear the specs are and how closely agents follow them. The announcement offers no measurements of that yet.
Code layers move to public beta
Figma also widened access to code layers, a feature that adds coding to designs on the canvas so prototypes can be more interactive. Figma announced it in private beta in June. At the India event it became a public beta, so many more users can try it.
Agents play a role here too. Users can ask the agent to turn a design into code, or to change part of the code while they iterate on a prototype. That blurs the line between a design file and a working piece of software, which is the direction Figma has been moving in for some time.
What Figma's design chief says about AI and code
Loredana Crisan, Figma's chief design officer, explained the thinking behind the release. She said Figma sees itself as the place where software teams decide what they build and ship, and that AI's biggest change is making it easier to build things.
She backed that with two numbers. By her account, about 17 times more lines of code are being produced today than in the past, yet releases have grown only about 1.3 times. Her reading is that much of that code was never meant to ship. It is written to explore what a product should be, which she called building to design. If that is right, design tools have more work to do, because more of the exploring happens in code and prototypes and not in static mockups. Those figures are Figma's own and come with no methodology in the announcement.
More tools can connect through MCP
Figma said its MCP now supports more than 20 new tools. Examples named at the event include Open Code, Dia, Grok Bot, and the Indian AI startup Sarvam. MCP, short for Model Context Protocol, is an open standard that lets AI tools connect to applications and data. In practice, it is how an outside agent gets access to what lives in Figma.
Broader MCP support fits the Specs idea. If many agents from many companies can reach a team's design context, the rules need to be written somewhere they can all read. Having a person re-explain the brand to each tool does not scale.
Payload joins Figma Make
Figma acquired Payload last year. Payload is an open-source headless content management system, meaning it stores and manages content separately from how that content is displayed. Figma is now integrating Payload with Figma Make, its product for generating dynamic websites and managing content directly. The company said Payload's downloads have grown sixfold over the last 12 months.
The pairing links two jobs that are usually split: designing a site and running the content that fills it. The announcement doesn't spell out every capability, so check Figma's own documentation for details on what the integration does and who can use it.
The event itself is part of the story. Figma opened its first office in India late last year, with teams for sales, partnerships, go-to-market, and product support. The company says India is its biggest market by users outside the United States, and that the platform works with 40% of the top 100 companies listed on the Bombay Stock Exchange. Holding its first Config event there signals how much weight Figma now puts on the country.
What this means for design teams
Written rules become more valuable: If agents read specs, then a clear, current spec becomes a working asset. Teams with vague or outdated guidelines will get vague results from agents.
Design systems get a new reader: Until now, design systems were written for people. Specs treats agents as a second audience, which changes how documentation should be written: precise, example-backed, and easy to parse.
Governance matters more: With agents acting inside Figma, someone has to own the spec, keep it current, and review what agents produce.
Open questions remain: The announcement does not give pricing, availability dates for every feature, or evidence of how reliably agents follow a spec. Treat the claims as promising but unproven until teams test them.
Frequently Asked Questions
What is Figma Specs?
Figma Specs is a Markdown-based editor in Figma where teams define their design language. The document serves as a reference for teams and AI agents, and it can reference canvas elements and embed FigJam diagrams, code layers, and designs.
Where does a Figma spec live?
It can be a standalone file or sit inside the canvas alongside the designs it describes.
How does Specs help AI agents follow brand rules?
Teams can bring design libraries, code libraries, and guidelines into one centralized system. Agents, both Figma's own and third-party ones, can read that system and follow its rules.
What are code layers?
Code layers add coding to designs on the canvas, so prototypes can be more interactive. They launched in private beta in June and are now in public beta. Users can ask the agent to turn a design into code or edit part of the code.
Why does Figma say AI is hard to control?
Because AI is non-deterministic, the same request can give different results. Figma says it is providing tools to steer agents toward more predictable outcomes.
Which tools does Figma's MCP support now?
Figma said it supports more than 20 new tools, including Open Code, Dia, Grok Bot, and Indian AI startup Sarvam.
What is Payload, and how does it connect to Figma?
Payload is an open-source headless content management system that Figma acquired last year. Figma is integrating it with Figma Make to generate dynamic websites and manage content directly.
