Niblet × Evidence-firstEvidence-first UI references, not mood boards
A reference is useful only if it answers a design question.
Mood-board tools (Gummble and the like) optimize for collecting inspiration. Niblet optimizes for answering questions: every screen is indexed by what it does, so “how do real apps handle a paywall after onboarding?” is a search, not a scroll. Collections feel productive; evidence ships product.
Set up
Search by job, not by app
The catalogue is captioned and embedded so retrieval works on what a screen accomplishes — onboarding, checkout, empty states — across hundreds of apps.
Inspect, don’t just collect
Every reference carries a full-resolution inspect URL, the source app, platform, and why it matched. You can defend the choice in review.
Close the loop
References plug into the agent workflow: brief → reference → implement → rendered check. The same evidence that inspired the work reviews it.
The workflow
- Ask
- Phrase the design question concretely.
- Retrieve
- Get 1–3 relevant screens, not 50 bookmarks.
- Build
- Implement against the evidence; verify by rendering.
Try the search on the home page — no account needed.
Read the docsQuestions? Contact the project. Legal: Terms · Privacy