There's a new design tool in open alpha called paper.design. It looks like Figma. It works like Figma. But underneath, every element you draw is real HTML and CSS. Not a proprietary vector format. Actual web code.
That sounds like a technical detail. It's not. It's a fork in how design tools will work from here on out.
👉 Two Different Bets
Figma stores your design as a proprietary node tree. It's optimized for visual fidelity and design system management. It's great for that. But when an AI agent needs to turn your Figma design into code, there's a translation layer. The agent has to convert Figma's format into HTML, CSS, and React. That translation is lossy. Developers working with the Figma MCP server report around 85-90% styling accuracy. Good enough to scaffold. Not good enough to ship.
paper.design skips that translation entirely. The design is already code. When an agent reads a paper.design file via its MCP server, it gets real HTML and CSS. No conversion. No guessing. The output quality is better because there's nothing to translate.
Figma's bet is: we're the best design tool, and we'll build bridges to code. paper.design's bet is: what if the design and the code were the same thing from the start?
🔥 Why This Matters for Founders
If you're a solo founder or a two-person team building something AI-native, this matters right now.
With paper.design, your AI agent can read from and write to the canvas. It has 24 MCP tools compared to Figma's 3. Your agent can sync design tokens from your codebase into the canvas. It can populate frames with real API data. It can generate responsive variations. It can convert a design to React and Tailwind and commit to GitHub directly.
With Figma's MCP, the agent reads context and generates code elsewhere. It's a one-way context bridge. paper.design is bidirectional.
For a founder who's vibe coding an MVP, that bidirectional flow means fewer steps between idea and shipped product. No exporting. No handoff. No "developer, please interpret this Figma file." The design is the code.
🧠 When Figma Still Wins
If you're a 15-person product team with a design system built over three years, paper.design isn't ready for you. Components, advanced prototyping, version history, branching, organization-level permissions. Figma has a decade of infrastructure that paper.design doesn't.
Code Connect, Figma's mapping layer between design components and code components, is a genuine quality multiplier. When it's set up properly, the MCP output improves dramatically. That investment compounds over time, especially for teams with mature design systems.
Figma also has the ecosystem. Plugins, community files, integrations with every tool your team already uses. paper.design is in open alpha. APIs can change. Features can break. Support is limited.
The choice isn't "which is better." It's "what stage are you at?"
⚠️ The Honest Take
paper.design is the most interesting design tool for AI-native workflows right now. But "interesting" and "production-ready" are different things. If you're building internal tools, prototyping fast, or working solo with high tolerance for rough edges, it's worth trying today.
If you have delivery deadlines and clients expecting polished output, Figma is still the safer bet. The MCP story is catching up, and Figma's recent Claude Code to Figma integration shows they understand the direction.
What you should pay attention to is the underlying question: should design tools store designs as pictures that get translated to code? Or should they store designs as code that gets rendered visually?
That question will define the next five years of design tooling. And it will determine which agencies can deliver AI-native workflows and which are still doing handoffs like it's 2022.
