For developers and AI agents

Put the library inside your editor.

A plain JSON API on every plan, and an MCP server so your coding agent can look up a real pricing table before it writes one.

MCP server

Add this to Cursor (.cursor/mcp.json), Claude Code or Codex:

mcp.json

{
  "mcpServers": {
    "uidiscovery": {
      "command": "node",
      "args": ["/path/to/uidiscovery/mcp/server.mjs"],
      "env": { "UIDISCOVERY_URL": "https://uidiscovery.com" }
    }
  }
}

Claude Code

claude mcp add uidiscovery --env UIDISCOVERY_URL=https://uidiscovery.com -- node /path/to/uidiscovery/mcp/server.mjs

Once connected, your agent gets a short guide to the library and these tools. Search results include preview images the model can actually look at, plus a link for every reference:

search_sectionsWebsite sections — hero, pricing, features, testimonials, FAQ, footer… Describe one section in plain language.

search_screensApp screens (platform ios) or web-app / dashboard UI (platform web).

search_flowsMulti-step journeys recorded from real products, with every step in order.

search_sitesWhole products by name, industry or style, with their section ids.

get_design_briefThe full agent-ready build brief (structure, tokens, font swaps, build steps, checklist) for a product slug or a section/screen id, with a preview image.

get_image_planThe reference's image style plus, per image slot, Unsplash / Pexels / Pixabay searches and a ready AI image-generation prompt.

get_design_tokensColour roles, palette and font stacks as JSON, ready for CSS variables or Tailwind.

The server tells your agent the workflow: search → look at the previews → get_design_brief → get_image_plan → build an original design → cite. There's also a design_with_references prompt: give it a brief and the agent researches references first, then designs something original.

REST API

Catalog lists return at most 25 records per request. Use offset and the response's nextOffset to load another page. List reads allow 120 requests per minute per connection; a 429 response includes Retry-After. Capture dates and age in days are reported where available, and older captures remain visible.

GET /api/v1/sections

curl "$UIDISCOVERY_URL/api/v1/sections?type=pricing&style=Dark&limit=3"

idOne section by id

qMatch site name, headings or copy

typehero, features, showcase, logos, testimonials, stats, pricing, how-it-works, integrations, faq, cta, blog, team, about, statement, content, footer, screen

categoryAI, Technology, Finance, Shopping, Lifestyle, Portfolio

styleLight, Dark, Minimal, Colorful, Serif, Mono

limit / offsetPaging, up to 25 per request

Each result includes the screenshot URL, palette, fonts, headings and a ready-made build prompt.

GET /api/v1/sites

curl "$UIDISCOVERY_URL/api/v1/sites?category=Finance&color=blue"

GET /api/v1/prompts

curl "$UIDISCOVERY_URL/api/v1/prompts"

Prompt results are original UIDiscovery build briefs with intent, tags and links back to the relevant reference patterns.

GET /api/v1/search

curl "$UIDISCOVERY_URL/api/v1/search?q=fintech+onboarding"

Search combines sites, sections and prompts. Explicit platform words (ios, website) and domains (fintech, education) boost matching references before ranking. When TYPESAFE_API_KEY is configured, JEV re-ranks the top of the fast shortlist by product-design relevance (results carry rankedBy and a cached flag); append ai=0 for deterministic fast ranking.

GET /api/v1/brief — build brief for one reference

curl "$UIDISCOVERY_URL/api/v1/brief?slug=linear"                  # whole page, Markdown
curl "$UIDISCOVERY_URL/api/v1/brief?section=linear-1"              # one section or screen
curl "$UIDISCOVERY_URL/api/v1/brief?slug=linear&format=json"       # structure, tokens, image plan + markdown
curl "$UIDISCOVERY_URL/api/v1/brief?slug=linear&part=images"       # imagery plan only

slugSite / app slug — a whole-page brief

sectionSection or screen id (alias: id) — a single-section brief; slug is optional

formatmd (default, text/markdown) or json

partimages — only the imagery plan (stock searches + AI generation prompts per slot)

The same brief powers the Build with AI button on every site and section page, the kit's PROMPT.md and the MCP server. It asks for an original build: new copy, your brand, free font stand-ins for licensed faces, and images from royalty-free stock or generation prompts — never the reference's own assets.

GET /api/v1/brief — map a product brief to references

curl "$UIDISCOVERY_URL/api/v1/brief?brief=calm+mobile+onboarding+for+a+finance+app"

Brief mapping uses JEV to choose a product journey and platform, then returns a ranked set of relevant sites, sections, and original build prompts.

GET /api/v1/flows

curl "$UIDISCOVERY_URL/api/v1/flows?track=onboarding"

Flow kits return an original three-step build sequence with the matching reference queries and prompt ids.