Neura Market
anydesign logo

anydesign

Free

Analyzes any image, URL, or Figma file and generates a structured `design.md` with the full design system, component inventory, and reconstruction notes — portable to v0, Lovable, Cursor or any AI builder. *By [@uxKero](https://github.com/uxKero)*

FreeFree tier
Inputs: image, url
Type
Open Source
Visit Website GitHub (40 stars)

About anydesign

anydesign is a Claude skill that transforms images, websites, and Figma files into structured, machine-readable design specifications. It outputs universal Markdown and W3C DTCG JSON that can be consumed by AI builders like v0, Lovable, Cursor, Bolt, Claude Code, or a human designer. The tool operates in two modes: full analysis extracts an entire design system (7-section spec, token inventory, optional WCAG report) while element mode isolates single components (e.g., navbar, 3D artwork) into focused rebuild prompts or token-grounded image prompts. Every inference carries a confidence marker (✅ ⚠️ ❓) and uses real hex colors—never invented tokens. Proven examples include rebuilding vercel.com from its design spec, recreating a retro audio player from a single captured element, generating consistent game asset packs, and producing IP-safe image prompts from art references. The project is open source and free.

Key Features

Full analysis mode: extracts a 7-section design.md spec + design-tokens.json (W3C DTCG) + optional WCAG accessibility report
Element mode: isolates single UI components or artwork and generates an element.md with rebuild prompt or token-grounded image prompt
Confidence markers (✅ ⚠️ ❓) on every inference to indicate certainty level
Output is universal Markdown and DTCG JSON, portable to v0, Lovable, Cursor, Bolt, Claude Code, or a human designer
No token invention: uses real hex colors and only reports what is observed
Supports images, website URLs, and Figma files as input
Includes CLI scripts for automation and batch processing
Open source with example run scripts and reproducible results

Pros & Cons

Pros
  • Vendor-neutral output works with multiple AI builders and human designers
  • Confidence markers reduce hallucinations and make inferences transparent
  • Element mode enables precise capture and reconstruction of individual UI components
  • Real-world examples demonstrate end-to-end reproducibility
  • Fully open source and free to use
  • Handles both visual design systems and generative art prompts
Cons
  • Requires Claude (or another AI that supports the skill) to execute; not a standalone application
  • Setup involves cloning a repository and configuring a Claude skill, which may have a learning curve
  • Output quality depends on the underlying AI's vision capabilities and prompt interpretation

Best For

Extract a full design system from a live website and rebuild it in an AI frontend builder (e.g., v0)Capture a single UI element (navbar, player, card) and recreate it as a working componentGenerate token-grounded image prompts for 3D art or illustrations while respecting IP boundariesCreate consistent game asset packs (character, pickup, prop) by extracting art direction from one key art

FAQ

What is anydesign?
anydesign is a Claude skill that turns images, websites, and Figma files into structured, machine-readable design specs (design.md and DTCG tokens) for use with AI builders like v0, Lovable, Cursor, Bolt, or a human designer.
Is anydesign free?
Yes, anydesign is open source and free to use. It is hosted on GitHub under the uxKero/anydesign repository.
What inputs does it accept?
It accepts images (screenshots, artwork), website URLs, and Figma file links.
What formats does it output?
Full analysis produces a design.md spec, design-tokens.json (W3C DTCG format), and optionally a WCAG report. Element mode outputs an element.md with code/prompt spec.