Figma Canvas | AI agents can now design directly in Figma files
Figma has opened its design canvas to AI agents, allowing tools such as Claude Code, Codex, and other MCP clients to create and modify design assets directly inside Figma files. Published on March 24, 2026, the update introduces the use_figma tool and skills, giving agents more access to design systems, components, variables, and team conventions.
Figma lets AI agents work directly on the design canvas
Figma's update focuses on a key limitation in AI-generated design: agents often lack the detailed product context that makes a design feel aligned with a real team's standards. Colors, spacing, typography, components, interaction patterns, and design-system decisions usually live inside Figma, not in a generic prompt.
For designers and product teams, the new workflow means AI agents can work closer to the source of truth. Instead of generating isolated mockups that feel disconnected from an existing product, agents can now create and update Figma files using the structure, variables, components, and conventions already defined by a team.
How agents work with the Figma canvas
The new use_figma tool allows agents to generate and modify design assets directly on the canvas. Figma says the workflow works through the Figma MCP server and can be used with MCP clients such as Claude Code, Codex, Copilot CLI, Copilot in VS Code, Cursor, Augment, Factory, Firebender, and Warp.
The tool complements generate_figma_design, which turns HTML from live apps and websites into editable Figma layers. In practice, generate_figma_design can bring an implemented UI back into Figma, while use_figma can then edit or create new assets using components, variables, and design-system context.
New workflow options for AI-assisted product design
The most important addition is skills. Figma describes skills as markdown-based instructions that tell agents how to execute workflows, which steps to follow, what conventions to respect, and what good output should look like inside a specific design environment.
For design teams, skills can make agent output more predictable. They can encode spacing rules, component usage, accessibility specs, design-token sync, component generation, layout conventions, and multi-agent workflows, turning team knowledge into instructions that agents can apply while building on the canvas.
The broader shift is toward code and canvas working in the same shared context. Designers still need to review hierarchy, component accuracy, accessibility, responsive behavior, brand fit, and final implementation quality, but agents can now help create and revise design assets without starting outside the product system.
Availability and beta access
Figma says the capability is currently available for free during the beta period and will eventually become a usage-based paid feature. The company also says the tool benefits from the security and reliability of the Figma MCP server while opening access to surfaces such as Figma Draw and FigJam through the Plugin API.
Figma is also expanding what agents can do on the canvas, with future work aimed at deeper native AI functionality, easier skill sharing, image support, custom fonts, and more parity with the Plugin API. For production teams, the best use case is controlled experimentation inside existing design systems before relying on agents for critical deliverables.
Daisuki's Take: What This Means for Designers
We see this Figma update as important because it brings AI agents closer to the real design environment instead of keeping them outside the workflow. When an agent can work directly on the canvas with components, variables, tokens, and team conventions, the output has a better chance of matching the product system instead of feeling like a disconnected mockup.
The strongest use case is controlled production support. Designers can use agents to explore interface changes, update assets, apply system rules, or prepare variations while still working inside the same Figma file. This can reduce repetitive setup work, especially for teams that manage large design systems or need to keep product screens consistent across multiple flows.
For professional work, the limitation is still review and responsibility. AI agents can help create and modify design assets faster, but teams still need to check hierarchy, component accuracy, accessibility, responsive behavior, naming conventions, and implementation quality. The best use case is treating agents as assistants inside the canvas, while keeping final design judgment with the team.
Sources and Recommended Links
- Agents, meet the Figma canvas | Figma Blog (Official)
- Introducing our MCP server: Bringing Figma into your workflow | Figma Blog (Official)
- Claude Code and Figma: Set up the MCP server | Figma Help Center (Official)
