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

Software design has always been a game of translation. You draw a box, tell a developer what the box should do, and hope the resulting code doesn't break the user experience. Figma’s latest moves suggest that this manual handover is finally starting to die. By opening up their platform to AI agents and moving their code layers feature into beta, they aren't just adding tools; they are redefining the canvas as a functional workspace for non-human logic.

The Shift to Agentic Workflows

For years, Figma was just a digital whiteboard. It was better than the competition because it lived in the browser and allowed for real-time collaboration. But the goal was always the same: create a picture of a product. With the introduction of specific frameworks for how AI agents interact with the platform, Figma is admitting that the future of design isn't just for humans looking at screens.

We are entering an era where design files need to be machine-readable in a way that goes beyond simple CSS exports. If an AI agent can understand the hierarchy, intent, and state of a design, it can start to perform tasks that used to take hours of tedious manual labor. We are talking about automated layout adjustments, accessibility auditing at scale, and even the generation of functional UI components based on high-level prompts.

Why Code Layers Matter

The transition of code layers into open beta is the most important part of this announcement for builders. Historically, design and code have lived in two different silos. Designers work in pixels; developers work in logic. Code layers bridge that gap by allowing the design file to hold the structural reality of the final product.

This isn't just about 'Export to React.' It's about maintaining a single source of truth. When code layers are integrated, the design isn't a lie that developers have to interpret. It becomes a blueprint that an AI agent can execute against. For founders, this means shorter development cycles and fewer 'lost in translation' bugs that eat up your runway.

The Founder’s Perspective

If you are building a startup right now, you need to look at this through the lens of efficiency. The old way of hiring three designers to churn out variants of the same landing page is over. These new features allow a lean team to leverage agents to handle the grunt work of design systems. You set the rules, the agents follow them, and the humans focus on the actual product strategy.

However, there is a catch. Skepticism is healthy here. We’ve seen 'AI-powered design' tools before that produced nothing but messy code and unusable layouts. The success of Figma’s agent strategy depends entirely on how well they handle the edge cases. If the agents can’t understand complex state changes or nuanced user flows, they’ll just become another layer of technical debt that your engineers have to clean up later.

Agents as Team Members

Figma is positioning these agents not just as plugins, but as active participants in the file. Think of them as junior designers who never sleep. They can organize your layers, check for consistency across a hundred screens, and ensure that your brand guidelines are being followed. This frees up your lead designers to actually think about the user, rather than spending their day renaming 'Frame 402' to 'Submit Button.'

For the crypto and AI builders I talk to, the takeaway is clear: stop thinking about design as a static asset. If your design environment isn't ready to talk to your codebase, you’re already behind. Figma is building the infrastructure for a future where software builds itself, and these agent frameworks are the first real step in that direction.

The Reality Check

We shouldn't get blinded by the hype. Just because Figma is defining how agents work doesn't mean the agents are ready to take over yet. There is still a massive gap between a 'generated design' and a 'good design.' Logic can be automated, but taste is much harder to codify. Builders should use these tools to accelerate their workflow, not replace their critical thinking.

The move to open beta for code layers is a signal that the tech is stable enough for public stress testing, but it’s still a beta. Expect friction. Expect the AI to hallucinate a few buttons that don't make sense. The goal for founders should be to integrate these tools slowly, finding the specific points in their pipeline where automation actually adds value instead of just adding noise.

What This Means for the Future

Ultimately, Figma is trying to own the entire lifecycle of a product, from the first brainstorm to the final line of code. By making their platform the default environment for AI agents, they are ensuring that they remain the center of the development universe. For those of us in the trenches, it means the tools are getting smarter, but the stakes are getting higher. You can build faster, but so can your competitors.

The real winners won't be the people who use AI to generate the most designs; they will be the builders who use these agent frameworks to create the most seamless transition from an idea to a working, shipping product. The distance between 'thought' and 'thing' is shrinking, and Figma just shaved off a few more miles.

The value of a design tool is no longer in what it lets you draw, but in how much friction it removes between the idea and the execution.

We are moving toward a world where the 'handover' doesn't exist. There is only the build. Figma’s bet on agents and code layers is a bet on that reality. It’s a smart move, but as always, the execution will determine if this is a revolution or just another set of features we eventually ignore.

Read the original at TechCrunch AI →