Figma adds native AI design agent

Figma added an AI design agent to its collaborative canvas this week, putting generative and automation tools directly inside the company’s core design surface. The product began rolling out on May 21 and is launching first in Figma Design, according to company materials. Figma said users can prompt the agent in natural language to generate new designs, edit existing work, automate repeated tasks and apply design-system guidance. The company said the feature is designed to work with the context already present in a team’s files, components and workflows. ### Why is Figma putting the agent on the canvas instead of beside it? Figma said in a March 24 blog post that “the canvas is now open to agents,” describing a system that can design directly in the workspace where teams already collaborate. The company said users can pair agents with “skills” that carry team-specific context, decisions and intent, rather than relying on a generic assistant operating outside the file. (cxotoday.com) The product page says the agent can apply a design system, automate bulk edits and give feedback while a designer works. That placement matters because the tool can act on the same frames, components and multiplayer context that human collaborators already use, according to Figma’s own description. ### What can the new agent actually do inside Figma Design? TechCrunch reported on May 20 that Figma’s agent accepts natural-language prompts to generate new designs, revise existing ones and automate iteration work. (figma.com) Figma’s help documentation and release notes describe related workflows that connect agents, MCP integrations and repeatable “skills” to move from product requirements or code into prototypes on the canvas. (figma.com) Figma has also been linking design more closely with code. In a February blog post about Claude Code to Figma, the company said developers could turn live interfaces from production, staging or localhost into editable frames on the canvas. That earlier work shows the company has been building toward agent-driven workflows that span design and development, not just image generation. (techcrunch.com) ### Where do OpenAI and Anthropic fit into the launch? Dataconomy reported on May 21 that Figma partnered with OpenAI and Anthropic for the in-house agent. Figma’s broader AI materials also reference MCP support, a protocol originally developed by Anthropic, as part of how outside systems and agentic workflows connect into the platform. (figma.com) Figma has not framed the launch as a standalone chatbot. Instead, the company’s public materials describe a design-specific agent that sits inside Figma’s own interface and uses the surrounding file context. That makes the OpenAI and Anthropic relationship look more like model and protocol infrastructure than a separate branded assistant, based on the company’s published product descriptions. (dataconomy.com) ### Why is Figma making this move now? Figma reported first-quarter 2026 revenue of $333.4 million, up 46% from a year earlier, according to reports published after its earnings. Coverage of the agent launch said the company is facing heavier competition from Canva, Adobe and newer AI-native design tools as it expands its own AI product line. (dataconomy.com) The launch also fits a broader product strategy visible across Figma’s recent releases. The company’s newsroom says Figma has evolved into an AI-powered platform, while recent release notes highlight live demonstrations of teams “co-designing with AI agents,” connecting design systems to code and shipping prototypes faster. ### What happens next? Figma said the agent is available during a beta period and will eventually become a usage-based paid feature. (theaiinsider.tech) Dataconomy and other reports said the rollout starts in Figma Design, with expansion planned across other Figma products later. Figma’s release notes and AI product pages are the next places to watch for rollout details, pricing changes and product-by-product expansion. (figma.com) The company has already said users will soon be able to publish and share custom skills across teams and organizations, adding another layer to how the agent is governed inside enterprise workflows. (figma.com 1) (figma.com 2)