Onlook Wants Designers to Ship Code Without Knowing It

The Macro: The Design-to-Code Handoff Is Still Broken

Every designer I know has a version of the same complaint. They design something in Figma. A developer rebuilds it in code. The result looks slightly wrong. The designer files a ticket. The developer says “that’s how CSS works.” Repeat forever.

This is a trillion-dollar industry running on screenshots and hope.

Figma owns the design tool market right now. They bought out the collaborative design space, survived an attempted acquisition by a large company, and remain the default for product teams everywhere. But Figma outputs design files, not code. The Figma-to-code pipeline is a cottage industry unto itself, with tools like Anima, Locofy, and Builder.io all trying to bridge the gap. None of them have fully solved it because they’re all translation layers. They take a static design and try to reverse-engineer what the code should look like.

The other side of the aisle has code-first tools. Cursor and Windsurf are AI-powered code editors that are excellent if you already think in code. Webflow lets you build visually but locks you into their platform. Framer is beautiful but opinionated. None of these tools let a designer open an existing React codebase and just start moving things around visually.

The Micro: A Visual Editor That Actually Writes to Your Codebase

Onlook is an open-source visual editor that connects directly to your React, Next.js, or Vue project. You open your project in Onlook, see your actual components rendered on a design canvas, drag things around, change styles, and every edit writes back to your source code in real time. No export step. No translation layer. The code IS the design.

The founding team is a two-person operation out of YC’s Winter 2025 batch. Daniel Farrell is the CEO, a designer by trade who spent over a decade in product design before becoming one of the first 100 employees at Bird, where he worked on global expansion strategy. He also co-founded Wanderlift, a long-distance ridesharing startup, and led growth at DIMO. Kiet Ho is the engineering half, with time at Amazon and ServiceNow building the kind of large-scale systems that make a tool like this possible. The combination of a designer who understands go-to-market and an engineer who’s shipped at scale is exactly what this product needs.

The traction numbers are hard to ignore. Over 22,000 GitHub stars. Hit number one on Hacker News. Was the top trending repo on GitHub, above DeepSeek. That level of developer and designer enthusiasm doesn’t happen by accident. The product also includes AI assistance constrained to your existing design system, which is a smart choice. Unconstrained AI generation produces garbage. AI that knows your design tokens and component library can actually be useful.

They offer a free self-hosted version and paid cloud plans, which is the right monetization strategy for a tool trying to become infrastructure. Get adoption first, charge for collaboration and hosting later.

The Verdict

I think Onlook is solving the right problem in the right way. The design-to-code gap isn’t going away because designers and developers use fundamentally different tools. The fix isn’t better translation between those tools. It’s one tool that serves both.

The biggest risk is scope. Supporting React, Next.js, and Vue means maintaining compatibility with three frameworks and their constantly shifting APIs. Cursor gets away with being a general-purpose code editor. Onlook has to deeply understand component trees, styling systems, and rendering pipelines for each framework it supports. That’s a lot of surface area for a three-person team.

At 30 days, I want to see designers at real companies using Onlook on production codebases, not side projects. At 60 days, the question is whether the AI features are generating components that developers actually keep or immediately rewrite. At 90 days, watch the GitHub contribution graph. Open-source tools live and die by their community. If external contributors are shipping meaningful features, Onlook becomes very hard to compete with. If it’s still just the founding team committing code, the open-source strategy is a distribution hack, not a moat.