Schema 2025: Design systems for a new era

Design systems get a refresh for the AI era

As AI tools and a broader range of contributors enter the product development pipeline, design systems are taking on a new role. Once primarily a consistency layer between design and production, they are increasingly becoming the translation layer that helps AI understand both design intent and code. At Schema 2025, Figma announced a set of updates aimed at making design systems more flexible, more performant, and more connected to the codebase.

The changes span four main areas: managing complexity in multi-brand setups, closing the gap between design tokens and code, integrating with agentic coding workflows, and opening up design to more contributors without sacrificing craft.

Scaling design systems without added complexity

Design systems naturally accrue complexity as they expand across products and platforms. Several announcements target that friction point.

Extended collections address a common pain point for companies running multiple products with distinct brand identities. Previously, variable-based theming worked well for straightforward cases but struggled to accommodate multiple brands within one system. Extended collections let design system authors release a whitelabeled version that designers can extend with their own themes, publish, and reuse. These extensions stay tied to the parent system, inheriting updates like new variables or color changes while preserving any explicit overrides. Extended collections become available in November.

Slots lift a long-standing restriction on Figma components. Historically, adding new layers to a component instance required either pre-authoring hidden layers or detaching the component entirely, breaking its connection to the design system. With slots, users can add layers within instances and specify exactly which instances a slot accepts, balancing flexibility with design-system compliance. Slots are open for early access applications.

Check designs is a new linter that automates the tedious process of matching raw values to their corresponding variables. After marking a design ready for dev or triggering Check designs via quick action, a custom model surfaces elements like variables and suggests the right one for the context. Users can review suggestions before applying them, then hand off to developers with confidence.

Underpinning these features is a significant performance rewrite. Figma has refactored the architecture and data models that power core design systems features. The result: updating variables or switching modes is now 30–60% faster, and heavy state swaps have dropped from 3500ms to 350ms, or from 2500ms to 450ms, depending on the complexity of the variable interactions and components involved. These performance improvements are live now.

Bridging design systems and code

With AI tools increasingly producing code, the relationship between design systems and the codebase has become critical. Two updates strengthen that link.

Code Connect UI removes a barrier to adoption for Code Connect, the feature that maps Figma components to their production code equivalents. The new UI lets teams connect Figma directly to GitHub repositories and relies on AI suggestions to quickly match the right code files to Figma components—no coding required. This flow is designed to reduce the setup and maintenance burden for design system teams. It is now rolling out to Organization and Enterprise customers.

The Figma MCP server, which brings Figma context into agentic coding workflows, has moved from beta to general availability. New capabilities include adding guidelines that tell AI models how to adhere to a design system, and parity between the remote and desktop servers. FigJam diagram support enables multi-step workflows and interaction wiring through agentic coding tools. MCP access is expanding to all users.

Design systems reach more of the team

As AI makes design more accessible, Figma's position is that more contributors don't have to mean lower quality—provided designers define the bar. Updates to Figma Make, the prompt-to-app tool launched at Config 2025, aim to bring production-grade design systems into AI-assisted prototyping.

Make kits allow teams to import Figma libraries directly into Make. The feature generates React code components and CSS files for styles and variables, then packages them for use in the tool. For teams keeping their design systems in code rather than Figma, npm package imports will accept React components—self-built or open-source—via public and private npm imports. Make kits are available for early access applications.

Variable tooling gets community-driven upgrades

Figma is rolling out a set of highly requested variable features over the coming weeks, all aimed at giving design systems more room to scale. The updates target variable import/export, collection authoring, and mode limits—three areas where the community has been asking for more flexibility.

Native import and export arrives

After years of relying on an open source plugin, Figma is adding native support for importing and exporting variables. The delay was intentional: the team wanted to wait for the Design Tokens W3C Community Group (DTCG) to finalize its 1.0 specification, ensuring that files moved between tools in a standard, interoperable format. With the spec now at 1.0, native import and export will be available in November.

A clearer authoring modal

Creating variables is getting a context boost. The authoring modal will soon display all subscribed collections directly, so designers don't have to switch views to understand their design system architecture. Users can also create variables inline rather than navigating through separate dialogs. To accommodate the larger interface, the modal will open full screen by default.

These authoring improvements are scheduled for November as well.

More modes for larger systems

For teams whose theming and branding needs outgrow the previous limit, Figma has increased variable modes: Professional plan users get up to 10 modes, and Figma Organization customers get up to 20. The change is effective now.

The broader theme across these changes is making design systems easier to both build and consume, particularly as teams lean on AI tooling to accelerate workflows. The increases target a common bottleneck—design system authors who need to manage complexity without sacrificing adaptability.

Paige Costello, VP of Product at Figma, leads the Editor group responsible for Figma Design. Previously, she was Head of AI and Core Product at Asana, and held leadership roles at Intercom and Intuit.