Give your Agent a Figma frame link—not screenshots, not a raw file dump. Skills first build a clean cached design package, then generate baseline Swift UIKit for Cursor/Claude/Codex. Enterprise-ready: plug your color/font/icon maps and base classes, full-screen coverage gates, optional API buffer against Figma 429s. Open source—you own the wiring for production APIs.
Hey Product Hunt 👋
I’m the maker of figma-ios-codegen.
What inspired this?
AI agents are great at writing UI—but feeding them screenshots or raw Figma dumps is noisy. Context blows up, tokens get invented, and you still don’t get your company’s design-system APIs.
Problem we solve:
Paste a Figma frame link (not a screenshot, not a giant JSON paste). Our Agent skills first build a clean, cached design package, then generate baseline Swift UIKit for Cursor / Claude / Codex.
How it evolved:
We split the pipeline on purpose—Phase 1 is deterministic data cleaning (no LLM); Phase 2 is skills + your host bindings (colors/fonts/icons/base classes). That made enterprise customization possible without forking skills.
Honest scope:
Baseline UI (layout/styles/mocks). Production APIs stay as TODO. Needs a Figma token. Open source under MIT.
Repo: https://github.com/mythkiven/fig...
Docs: https://github.com/mythkiven/fig...
Release: https://github.com/mythkiven/fig...
Happy to answer questions about bindings, coverage gates, or why we don’t ship an MCP server.
Report
A native SwiftUI output option alongside the UIKit baseline would be huge, since a lot of teams are moving off UIKit and it'd be annoying to maintain two pipelines. Could the skill support a SwiftUI flag or a separate generator so we can pick per project
Report
love the skills-first approach—caching a clean design package before generating Swift keeps the token usage sane and the output actually usable. the plug-in points for our color and icon maps are exactly what we needed instead of fighting an opinionated codegen.
Report
finally something that skips the screenshot detour and just hands the agent the actual frame, the cached design package idea is the part i wish more tools copied
Report
Reviews
No reviews yetBe the first to leave a review for figma-ios-codegen
A native SwiftUI output option alongside the UIKit baseline would be huge, since a lot of teams are moving off UIKit and it'd be annoying to maintain two pipelines. Could the skill support a SwiftUI flag or a separate generator so we can pick per project
love the skills-first approach—caching a clean design package before generating Swift keeps the token usage sane and the output actually usable. the plug-in points for our color and icon maps are exactly what we needed instead of fighting an opinionated codegen.
finally something that skips the screenshot detour and just hands the agent the actual frame, the cached design package idea is the part i wish more tools copied