The Figma Agent Is Here. Here's What You Can Actually Do With It Today.
Figma's new on-canvas agent just shipped, and it's not just an autocomplete tool. Here are six concrete ways designers can use it right now to cut busywork, explore directions faster, and make their work more expressive.
Figma just launched its design agent, a collaborator that works directly on your canvas alongside you. The pitch is not "AI does the design for you." It is "AI handles the tedious parts so you can stay focused on the thinking."
That distinction matters. For designers who are starting to build with AI, whether that is prototyping vibe-coded projects or just trying to move faster inside Figma, the agent opens up a handful of genuinely useful shortcuts right now. Figma's own guide lays out six of them. Here is what they actually mean for your workflow.
Your feedback loop just got a lot shorter
The first two use cases are both about feedback, and they might be the most immediately useful for anyone shipping real work.
The first is transcript annotation. If you use a note-taking tool like Granola, Notion, or Google Drive, you can connect it to the Figma agent through the Connectors menu (the plus icon in the agent's chat box). Share a meeting transcript link and prompt the agent to annotate your designs based on the crit feedback in that transcript. It maps comments to the relevant frames on the canvas so you do not have to do that translation yourself after a long meeting.
The second is comment summarization. Select frames that have comments on them and ask the agent to summarize the feedback and suggest updates. There is even a community skill, a reusable instruction set you can trigger with a slash command, called /comments-do-try-consider that organizes feedback into three buckets: things to do, things to try, and things to consider. If your Figma files regularly fill up with scattered sticky notes and comment threads, this alone might be worth the exploration.
Custom shaders without writing a line of code
Shaders are code-driven visual effects, things like liquid metal, glass distortion, grain, or holographic textures. Historically, using them meant finding a developer or digging into code. Now you can describe the effect you want in plain language and the agent builds it. A grainy vintage look on a photo, a glass blur on a card, a subtle noise texture across a background, these are all fair game.
If you are new to the concept, there is a /shader-starter skill designed to ease you in. You can also pipe in a whole folder of images from Google Drive and ask the agent to apply the same shader across all of them in one go. For anyone building a branded app or a design system with a strong visual personality, this is a meaningful shortcut.
Exploring directions without burning a day
One of the most time-consuming parts of early design work is building multiple directions to compare. The agent can generate several layout variations of a screen or feature from a single prompt, so you can put real options side by side on the canvas quickly. If you already have a live website or app, the Figma Chrome extension lets you pull the latest version into a file in one click, giving the agent something real to riff on.
This is worth pausing on for designers building their own apps with AI. The exploration phase, where you are figuring out which layout actually works, is usually the part that slows everything down. Being able to spin up three or four directions at once and compare them visually is not a small thing.
Animation without leaving Figma
The source article also mentions animation as one of the six workflows, though the detail on this is lighter. The idea is that you can describe motion behavior to the agent and have it applied to your designs, keeping the work inside Figma rather than exporting to a separate tool. This is still worth watching as the feature matures, because the gap between a static design and something that feels alive in a prototype is often where non-engineers get stuck.
Skills are the feature worth learning first
Across all six use cases, the concept of skills keeps coming up. A skill is a saved, reusable set of instructions for the agent, shareable with your team or pulled from the Figma Community. You trigger them with a / command. Think of them as macros built in plain language, not code.
If you are going to invest time in the Figma agent, learning how skills work is probably the highest-leverage starting point. The community library already has skills for shader work, comment organization, and more. You can also build your own to capture workflows you repeat often.
A grounded read on where this sits right now
The Figma agent is genuinely useful for the tasks above. It is not a replacement for design judgment, and it is not going to make decisions for you. The annotation and comment features depend on how well your transcripts and notes are structured. The shader outputs will need refinement through prompting, not just a single command. And the multi-direction generation is only as good as the brief you give it.
None of that is a reason to wait. These are real tools that work today, and the best way to understand what the agent can and cannot do is to try it on a project you already care about. Start with the feedback workflows if you work collaboratively. Start with shaders if you want something visual and immediately satisfying. Either way, the canvas is a good place to find out.
Sources
About VibeLab
VibeLab is a community for vibe coders. Discover vibe coded projects from around the globe, and post and host your own.