Figma drops native AI agents directly onto the canvas

Designing UI usually means spending half your day on pure digital grunt work: tweaking corner radiuses across dozens of screens, swapping component states, or generating yet another slight variation for client review.

Figma is tackling that headache head-on. On May 20, the company introduced a native AI agent directly inside its collaborative canvas, meaning you can now delegate design tasks without ever leaving your file.

Multiple agents working the artboard

Rolling out first to Figma Design before expanding across other products, the new agent is designed around a simple interaction model: plain text prompts.

Instead of fiddling with individual layers manually, you can tell the agent to spin up brand new layouts, edit existing designs, or generate multiple design iterations automatically. Figma also allows users to fire up multiple agents at once, letting different agents tackle separate tasks across the canvas in parallel.

Under the hood, the company claims the agent actually understands design context and hierarchy because it runs on models fine-tuned specifically for design workflows.

As Figma’s Chief Design Officer Loredana Crisan put it, as building software gets easier, the hardest part becomes setting direction. The goal here is to let teams collaborate with agents on multiplayer files to visualize edge cases and iterate concepts together without getting bogged down in tedious pixel-pushing.

Defending territory while bridging to code

The launch builds on Figma's recent push to bridge design and engineering. Over the past few months, the company partnered with Anthropic and OpenAI to support CLI tools like Claude Code and Codex directly alongside its software.

Figma is fighting on multiple fronts against giants like Canva and Adobe, as well as newer AI challengers like Flora, Krea, and Dessn. To stay ahead, Figma acquired node-based tool Weavy last year and folded in new image-editing features.

So far, the moves seem to be paying off: Figma posted $333.4 million in revenue for Q1 2026, marking a 46% jump year-over-year despite broader industry fears that AI might diminish software design seats.

Why it matters

For designers and product teams, this shifts the job description from manual execution to art direction.

Instead of spending hours manually drafting empty states, error banners, or mobile responsiveness variations, you can spin up parallel agents to generate those baseline options in seconds. The fact that it lives natively inside the multiplayer canvas means zero friction: no jumping between external web generators and copy-pasting SVGs.

Strategically, Figma is doing the smart thing by focusing on design-tuned agents that respect layers rather than slapping on a generic diffusion model. Pure image generation has always been useless for real interface engineering; understanding auto-layout, vectors, and design tokens is where the real moat lies.

The real test will be execution discipline. If these agents break component properties or leave messy, unorganized frames behind, designers might spend just as much time cleaning up AI clutter as they save in drafting. But if Figma nails layer cleanliness, this could change everyday UI workflows permanently.