Figma
The collaborative interface design tool
4.9•1.6K reviews•14K followers
The collaborative interface design tool
4.9•1.6K reviews•14K followers
A collaborative design tool for creating user interfaces, mobile apps, and websites with a wide range of features, including vector editing tools, prototyping, and version control, all in a cloud-based platform.
This is the 34th launch from Figma. View more
Figma Agent
Launched this week
Figma's agent lives right on the Design canvas and works with your real components, variables and team files, not flat images. Generate layout directions, bulk-edit across screens, turn comments into changes and build prototypes from a prompt. Pull context from Notion, Slack, GitHub and Linear via MCP, build your own plugins and shaders, and save repeatable workflows as skills your team runs with "/". Stay in flow from idea to review.
Free Options
Launch Team
Miro
Sketch
Canva
Adobe XD
Hey Product Hunt 👋
Hunting this one because it's the first AI design tool I've seen that doesn't make you leave your real work behind.
Most AI design tools hand you a nice-looking image and then you rebuild it by hand in Figma. Figma's agent skips that step. It works inside your file, with your components, tokens and auto layout, so what it makes is something your team can actually ship.
What stood out to me:
💠 Explore directions fast - several layout options or a dark mode version, all built from your own design system
💠Kill the busywork - bulk edits, realistic copy and images, components applied at scale
💠Real context - paste a Figma file link, search the web, or pull in PRDs and research from Notion, Slack, GitHub or Linear
💠Build your own tools - ask the agent for a plugin or shader and it shows up in a new Tools tab
💠Skills - turn a workflow you repeat (a crit prep, an accessibility audit, a brand voice pass) into a skill anyone on the team runs with "/"
The real story is staying in flow, so no exports, no tool switching, no re-prompting from scratch every time.
What's the first repetitive task you'd hand off to it? 👇
@kate_ramakaieva the figma file link and web search in one place sounds useful how well does the agent handle bigger design tasks?
With tools like these, I am thinking about whether people will be learning their traditional use. Remember how I was learning Figma 5 years ago to complete a job assignment. (I was watching Figma tutorials at 2x speed on YouTube to make it to the deadline). And this probably will not be needed anymore – the usage of the tools as well as those tutorials. Just prompting.
I am missing those times, to be honest, because I was earning that feeling that I own and learn. :D
@busmark_w_nika yeah, also remember that time, but still it works much better with that knowledge base and experience 🙌
Very cool! I guess the question we are all wondering 😸: How is the metered pricing on this? Can I connect with my chatgpt200 subscription and use that to drive the AI?
I really like the bulk edit functionality through multiple screens. Is there any way to see all changes that will be made and have an option to undo all at once?
Btw, Congratulations @zoink and @brendan_o_driscoll 🎊
Keep Hunting cool Product's @kate_ramakaieva 😎✌️
if I prompt an edit on a frame that uses components from our library, does the instance stay linked or does it get detached? asking because detached instances are the thing that quietly breaks a design system.
congrats on the launch!!🚀
as a FE engineer side, how do i find out what changed? bulk editing twenty screens sounds great for the designer, but i'm the one who has to work out which of those actually need code.
right now thats me eyeballing the file against what's already shipped does the agent leave a trace of what it touched, or is it just a new version of the file.
"turn comments into changes" is the feature I'd actually want to stress-test before trusting it on a real team file. a single clear comment like "make this button bigger" is a fine instruction to automate. but a lot of design comments are proxies for an unresolved disagreement, a PM and an engineer leaving different comments on the same element because they want different things, and the discussion that's supposed to happen between those two comments is the actual work. if the agent just applies both literally, it can make the file look resolved when the disagreement never actually got talked through. does it flag comments that look like they're in tension with each other, or just execute whatever's attached to the component