AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Figma announces new features to define how agents work on its platform

Collected Oct 15, 2026

Figma unveiled a set of new capabilities at its first Config event in India, centered on a Markdown-based editor called Figma Specs that lets teams write down their design language in a simple document. The announcement covers three connected pieces: the Specs editor itself, a public beta release of code layers, and an integration with Payload, the open-source headless content management system Figma acquired last year. Code layers, which add coding to designs on the canvas for more interactive prototypes, was first announced in private beta in June; at the India event the company is opening it to public beta so more users can test it.

Specs is the centerpiece. It works as a text editor, but users can reference other Figma elements inside it, including items on the canvas, embedded FigJam diagrams, code layers and designs. A spec can live as a standalone file in Figma or sit alongside content on the canvas. Teams can also pull in broader context such as design libraries, code libraries and guidelines, giving a design system a single place to manage its rules. The stated goal is that both Figma's own agents and third-party agents can consult this centralized system and follow a brand's design rules more reliably when generating with AI.

That connects to a broader theme: Figma made its agent on the canvas available to a wider user base in the past week, and Specs is best read as a steering mechanism. AI systems are non-deterministic, meaning the same prompt can produce different output each time. Structured, machine-readable context is one of the more practical ways to constrain that variance, which is why a Markdown format makes sense here — it is plain text, diffable, easy to version and easy to feed to a model as context. In practice, a design spec becomes something closer to a prompt artifact than a static documentation page: a contract that both humans and agents read. A likely trade-off is maintenance. A spec that drifts from the actual canvas or code library is worse than no spec, because agents will faithfully follow stale rules.

The company also said its MCP now supports more than 20 new tools, among them Open Code, Dia, Grok Bot and Indian AI startup Sarvam. MCP, the Model Context Protocol, is a widely adopted open standard for connecting AI assistants to external tools and data. Supporting a list of clients through it means Figma's context — including the design rules captured in Specs — can be surfaced inside whichever assistant a team already uses, rather than only inside Figma. The inclusion of Sarvam is notable given the venue; the integration list leans toward tools with traction in the Indian market.

On the Payload side, Figma is integrating the headless CMS with Figma Make, its product for generating dynamic websites, so teams can manage content directly. Figma said Payload's downloads have grown sixfold over the last 12 months. Headless CMS architecture separates content storage and delivery from the presentation layer, which pairs naturally with AI-generated front ends: the model can build the page while content stays in an editable repository rather than being baked into generated markup. For developers, that is a meaningful distinction, because regenerating a site does not mean losing editorial work.

Figma's chief design officer, Loredana Crisan, framed the shift around build capacity rather than release volume. She said there are 17 times more lines of code being produced today than in the past, while releases have grown only 1.3 times, and argued much of that code is not necessarily built to ship but built to discover what the product is — built to design. Read plainly, that is a claim that AI has inflated the volume of exploratory code far faster than the volume of shipped products, and it implies a market for tools that make the exploratory phase legible and reusable. That is roughly the bet Specs and code layers represent.

Code layers let users ask the agent to turn a design into code, or to modify part of the code to iterate on a prototype, which blurs the line between the design canvas and a working artifact. The public beta widens the pool of testers, and the feature's usefulness will depend on how cleanly generated code maps to a team's existing components and conventions — which is again where a spec and imported code libraries matter.

The event also carries an India-specific angle. Figma opened its first office in the country late last year, staffing sales, partnership, go-to-market and product support roles. The company said India is its biggest market by users outside the United States, and that its platform is used by 40 percent of India's top 100 companies listed on the Bombay Stock Exchange. The Config event is the company's first in India.

Why it matters: design and engineering teams get a durable place to encode brand rules that AI agents can consume, which could reduce the rework that comes from unpredictable generated output. Developers and product teams using Figma Make or a headless CMS should watch how Specs, code layers and the Payload integration fit into existing pipelines, and whether the spec stays authoritative as designs change. Teams evaluating agent-based workflows will likely treat Specs as a governance layer rather than a documentation artifact — though that is inference from the stated features, not a commitment Figma has made.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Figma is also releasing its code layers feature in open beta.