

Point your AI agent at any website. Get back a complete design breakdown β colors, type, spacing, and the reasoning behind every decision β ready to use in your next build.
Taste Lab is a design analysis tool that reverse-engineers the visual DNA of any website. Point it at a URL, and it returns a complete design breakdown β colors, typography, spacing, shadows, and the reasoning behind every decision. The output comes as two files: a .md design map and a .json token set, ready to feed into any AI-powered build tool.
Taste Lab runs four specialized AI agents in sequence. The first extracts raw measurements (every color, weight, spacing value, radius, and shadow). The second detects systematic rules from those measurements. The third infers the designer's taste β the deliberate trade-offs behind each choice. The fourth acts as a quality gate, filtering out sloppy reasoning and validating the output before writing the final files.
Every run produces a Design Map β a complete token set covering 20 measurement categories β and a Taste DNA section with four principles. Each principle includes the Trigger, Decision, Reason, Evidence, and Trade-off behind one design choice. At least one principle is a Restraint, explaining what the designer chose not to do.
No approximations. The output cites exact px, hex, ratio, and DOM values for every measurement. For example, a page background might be #08090A (near-black, not pure black), with a text primary of #F7F8F8 (nearly white, cool). Spacing is referenced to a base unit (e.g., 8px), and every shadow includes its full rgba and offset values.
Taste Lab doesn't just tell you what a site looks like β it tells you what the designer gave up to make it look that way.
Most design analysis tools stop at color palettes and font stacks. Taste Lab goes further by extracting the reasoning behind each choice β the trade-off that makes a design intentional rather than accidental. A principle like "Brand lives in white, not in color" explains why a near-black site uses white as its accent instead of a brand color, and what that choice sacrifices (e.g., the cheap feel of a template). This turns a static token set into a reusable design philosophy.
You're building AI tools that need to generate or adapt visual interfaces, and you want them to learn from real design decisions rather than random style guides. Also worth a look if you're a designer who wants to audit your own work with the same critical lens β or if you're reverse-engineering a competitor's site and need more than a color picker.
Other tools you might consider
Loading commentsβ¦
Maker
indie_inkwell
Visit Website
tastelab.xyz
Project Info
Product Keywords
Compare with
Achievement