Figma has released a design agent that runs directly on the canvas and in the left rail, positioned as a native alternative to the third-party agents the company opened its canvas to earlier this year. The company frames the launch around a rejection of two false choices it says are emerging as teams adopt agentic tooling: speed versus precision, and AI generation versus direct manipulation.
What the agent is, and how it differs from MCP
The agent is fine-tuned for editing Figma files, so outputs are tailored to a team's design context and remain objects you can manipulate directly. It carries deep context on components, tokens, standards, and best practices — legibility that Figma says is not possible for outside tools. Because it lives on the canvas, there is no separate setup or context switching, and you can start a prompt from any design layer, run several prompts in parallel to explore ideas at once, and keep editing while the agent iterates.
The distinction from Figma's other AI surfaces is scope rather than replacement. The MCP server and use_figma are the route for moving work between code and the canvas. The agent is the tool for working inside the file: it is built in and has additional context about your design system. Figma describes the split as using the agent while you're on the canvas and MCP to pull code in or push designs back out.
Exploration and iteration
To go wide, the agent generates distinct stylistic approaches to the same problem — multiple checkout flows tuned to different business goals, for example, or three alternative information architectures. To go deep, you pick a direction and ask it to iterate, compare implementations, and rethink existing designs while staying aligned with the system. It draws on your most frequently and recently used components as a starting point, and you can steer it further by selecting a specific library and @-mentioning tokens, variables, and components.
Figma notes that once a direction is chosen, going hands-on is often faster and more token efficient than prompting toward an exact result. The agent also slots into handoffs with Figma Make in either direction: generate design layers in Figma Design to clarify flows, states, copy, and structure, send them to Make for code layers that clarify behavior, then bring the result back. Teams can build simultaneously in the same workspace throughout.
Bulk edits and design system upkeep
The agent's second role is absorbing precision work that requires context and doesn't scale by hand: renaming variables for consistency, swapping one component across many screens, repeating a padding change through an entire flow, or populating many frames with realistic content. It also applies design systems across designs as it goes.
Typical requests include updating typography across a file, replacing lorem ipsum copy and imagery across a grid, setting all chip components to their active state, and converting screens to dark mode without manually reworking fills and contrast. For design system maintainers, the agent can bulk-update descriptions, tags, and use cases across libraries, standardize naming conventions, document components with all their states and variants, and take direction from examples of the team's own work.
Keeping code and canvas in sync
Work can start in code and come into Figma through code-to-canvas to iterate or apply the design system, then return via the Figma MCP server without losing fidelity. The agent sits inside that handoff to maintain momentum as work moves between tools.
Putting accumulated feedback to work
Feedback tends to pile up across comments and files — crit notes, stakeholder reactions, open questions. Because the whole team works in the same file, the agent already holds that context, which Figma says makes working through feedback feel like thinking out loud with someone already in the room rather than briefing a new collaborator. It can summarize feedback, identify themes, and turn input into next steps; model how a given stakeholder, such as a VP focused on revenue, might react to pressure-test a design; distill a long comment thread into an actionable plan; or help sharpen a direction before a crit.
Availability and roadmap
The agent rolls out gradually in beta over the coming weeks, with a waitlist for early access; joining the list does not guarantee access, and eligible users will receive an email when they're in. During beta the agent won't consume credits — AI credits apply at general availability. It will be available to Full seat users on Professional, Organization, and Enterprise plans, while Collab and Dev seats can use it in drafts. Starter, Education, and Government plans are not included.
Figma says its near-term focus is improving design system support, refining the user experience, expanding the agent's ability to search across your files, and adding customization.



