Get This Tool
Taste Lab
Pricing
- Model
- Free
Summary
Every design token scraper gives you the hex and the font size — none of them tell you why the designer chose near-black over pure black, or what they gave up to get there. Taste is an agent pipeline that extracts that reasoning, not just the spec sheet.
Point Taste at a URL and four chained agents work through the page in sequence: one pulls raw measurements across 20 categories, the next finds system-level rules, the third infers deliberate trade-offs, and a final 'observer' agent runs an anti-slop filter before writing output. The result is two files — a .md and a .json — that carry both the token set and four named design principles, each with a trigger, decision, reason, evidence, and trade-off. This makes it useful when you need a coding agent to replicate a visual language rather than just match numbers. The ceiling appears when the site's design logic is implicit or inconsistent — the pipeline infers intent, and on a messy codebase that inference will be wrong.
Bottom line: Pick Taste when you need to feed a coding agent the reasoning behind a design system, not just its values — but if the target site has no consistent design logic, the 'taste principles' the pipeline produces will be plausible-sounding noise.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Four-agent sequential pipeline with a built-in quality gate that runs anti-slop validation before writing output, which means the files your coding agent receives have been filtered once already rather than passed through raw.
- Output includes explicit trade-off documentation for each design principle — not just what was chosen but what was deliberately left out — so agents can make consistent calls on components the original spec never addressed.
- Writes both .md and .json on every run, so the same extraction drops into Cursor rules, Claude context files, or any tool that reads structured markdown without a conversion step.
- Self-hosted and free with no paid-only features mentioned in the docs, which means you can run it against internal or client sites without routing screenshots or DOM data through a third-party service.
- Requires at least one restraint principle per run, which forces the pipeline to document negative space in the design — the deliberate absences that pure token extraction misses entirely.
Cons
Sign in to edit- The pipeline infers designer intent from DOM measurements and visual patterns. On sites built from mismatched component libraries or accumulated patches with no governing system, the four 'taste principles' it produces will be confident and wrong — teams shipping agent context from those runs will be baking hallucinated design logic into their rules files.
- There is no API surface. The tool is a CLI-installed skill, not a service you call programmatically, which means teams who want to run extraction as part of a CI pipeline or trigger it from another agent workflow have to wrap it themselves or abandon it for a tool with an HTTP interface.
- The four-principle output structure is fixed. Design systems with more than four governing decisions — or ones that need principles organized by component type rather than by abstract choice — hit a structural ceiling that the tool does not expose a way to override, at which point teams either post-process the JSON manually or move to a custom extraction prompt they maintain independently.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- CLI, IDE extensions
- API Available
- No
- Self-Hosted
- Yes
- Last Updated
- 2026-06-21T14:18:59.752Z
Best For
Who it's for
- Design-aware AI agents
- UI/UX analysis workflows
- Replicating visual language across tools
What it does well
- Reverse-engineer design systems from live sites
- Supply consistent design context to coding agents
- Generate .md and .json files for Cursor, Claude, or Gemini rules
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Compare Taste Lab
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is Taste Lab free?
- Yes — Taste Lab is fully free to use. There is no paid tier.
- Is Taste Lab open source?
- Yes. Taste Lab is open source.
- Can I self-host Taste Lab?
- Yes. Taste Lab supports self-hosting on your own infrastructure.
- What platforms does Taste Lab support?
- Taste Lab is available on: CLI, IDE extensions.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Taste takes a URL and runs it through a four-step agent pipeline that terminates in two artifacts: a `{domain}.md` and a `{domain}.json`. The first agent extracts measurements — every color, weight, spacing value, radius, and shadow, cited with exact px, hex, or ratio. The second detects 5–8 system-level rules, each with an evidence line and a stated design goal. The third infers four ‘taste principles’, each carrying a trigger, decision, reason, evidence citation, and explicit trade-off. The fourth agent acts as a quality gate: it runs an anti-slop grep, validates the JSON, and discards anything that didn’t survive scrutiny. At least one of the four principles is required to be a restraint principle — a documented choice about what the designer left out.
The differentiating claim is in what the output carries beyond tokens. A spec sheet tells an agent the border radius is 6px. A Taste DNA principle tells it that 6px was chosen for primary cards, that 2px was reserved for micro-elements, and that the decision trades warmth for precision — so when the agent encounters a component the spec never covered, it has the reasoning to make a consistent call rather than defaulting to its training data’s average.
The tool fits inside workflows where you are supplying design context to Cursor, Claude, or Gemini via rules files — the .md output is structured for that handoff. It is self-hostable, free, and the vendor describes it as a GitHub-installable skill with no paid tiers. Where it breaks is on sites with no coherent design system: the pipeline is built to infer intent, and on a site assembled from mismatched libraries or iterative patches, the principles it produces will be internally consistent but factually wrong about what the designer intended. Teams working against those targets will need to treat the output as a starting draft and edit before committing it to agent context.