Liftit

Lift any site's design. Paste a URL, get the code.

npx liftit https://stripe.com

Liftit opens a real browser, extracts every computed style, animation, and interaction from a live website, and converts it into code you can use — Tailwind configs, React components, shadcn themes, or CSS variables. Works on any site regardless of framework.

Install

npm install -g liftit
npx playwright install chromium

What you can do

Extract a full design system

liftit https://stripe.com

Captures DOM, computed styles, screenshots at 8 breakpoints, fonts, images, animations, hover states, scroll behaviors, and interactive elements. Produces a complete design system spec.

Generate a React component from any site

liftit component https://linear.app header
liftit component https://vercel.com hero --name VercelHero

Outputs a working Header.tsx + Header.module.css with exact styles, hover states, responsive breakpoints, and animations. Paste it into your Next.js project.

Make your component match a target

liftit match https://linear.app header --file ./src/Navbar.module.css

Reads your existing CSS, extracts the target, and tells you exactly what to change:

.nav -> root (nav)
  CHANGE  padding: 0px 24px -> 12px 40px
  CHANGE  border-radius: 5px -> 10px
  ADD     backdrop-filter: blur(35px)

Writes a patched CSS file ready to replace yours.

Export to your framework

liftit export tailwind          # tailwind.config.ts
liftit export shadcn            # globals.css + tailwind.config.ts (light + dark mode)
liftit export css-variables     # design-tokens.css with :root custom properties
liftit export all               # everything

Quick token extraction

liftit tokens https://cal.com

Colors, typography, spacing, shadows in ~5 seconds. No full pipeline needed.

Scan with multi-page crawling

liftit scan https://stripe.com --crawl

Discovers and scans up to 20 internal pages. Supports auth cookies and headers for protected sites.

Claude Code Integration

liftit setup

One command registers the MCP server with Claude Code. Then talk naturally:

• "Extract the design system from this URL"
• "Generate a React component from Stripe's pricing section"
• "Make my navbar CSS match Linear's header"
• "Give me a Tailwind config from the tokens you extracted"

15 MCP tools available — extraction, component generation, matching, export, validation, and browser interaction.

How it works

1. Scan — Playwright captures the full DOM with 76 computed CSS properties per element, screenshots at 8 viewports, all fonts/images/SVGs, API calls
2. Analyze — CSS architecture detection (BEM, Tailwind, Modules), component detection via 4 strategies (semantic roles, structural patterns, style clustering, class names)
3. Motion Capture — Scrolls the page recording style changes at every position across 3 viewports. Hovers interactive elements frame by frame. Captures Web Animations API data
4. Interactions — Catalogs navigation, toggles, modals, dropdowns, forms, scroll behaviors
5. Scroll-Interaction Mapping — Connects animations to interactivity. Discovers which scroll animations produce or remove interactive elements
6. Distill — Compresses raw motion data into compact animation specs with from/to values, triggers, durations
7. Assets — Copies fonts, images, SVGs into your project with @font-face CSS generation
8. Synthesize — AI reads all data and produces structured design system JSON + human-readable spec
9. Validate — Pixel-level comparison (pixelmatch, 0.1 threshold) with DOM and interaction verification

Steps 2-4 run in parallel.

CLI Reference

liftit <url> [output-dir]                         Full extraction pipeline
liftit component <url> <name> [--output dir]      Generate React component
liftit match <url> <name> --file <css>            Diff your CSS against target
liftit export <format> [--input dir]              Export tokens
liftit tokens <url>                               Quick token extraction
liftit scan <url> [--crawl]                       Scan only
liftit validate <url> [--rebuild] [--site]        Pixel validation
liftit mcp                                        Start MCP server
liftit setup                                      Register with Claude Code

Flags

--crawl                Multi-page crawling (up to 20 pages)
--auth-cookie "k=v"    Auth cookie for protected sites
--auth-header "K: V"   Auth header for protected sites
--output <dir>         Output directory
--input <dir>          Input directory (for export)
--name <Name>          Custom component name
--file <path>          Your CSS file (for match)
--map "a=b,c=d"        Manual class mapping (for match)
--rebuild-url <url>    Dev server URL for validation

Output

output/
  design-system.json        Structured tokens + components
  design-system.md          Human-readable rebuild spec
  scan-result.json          Raw DOM and computed styles
  analysis-result.json      Component detection, CSS architecture
  motion-distilled.json     Compact animation specs
  scroll-interactions.json  Animation-to-interaction links
  interactions.json         Interactive behavior catalog
  screenshots/              Reference screenshots (8 breakpoints)
  assets/                   Downloaded fonts, images, SVGs
  export/                   Framework configs (Tailwind, shadcn, CSS vars)
  components/               Generated React components

Requirements

• Node.js 18+
• Playwright (npx playwright install chromium)

License

MIT