One search, many stores, no tabs.

What happens when you search

Results arrive in waves, so you see something useful right away and more fills in as it lands.

1. 1

   Instant catalog waveabout half a second

   Your words go to the Shopify Global Catalog through the Universal Commerce Protocol (UCP). It answers with live listings from thousands of independent stores — title, price, stock, and the store's own product link.

2. 2

   Big-store wavea few seconds later

   While you look, a second search asks Google Shopping (through an Apify actor) about big-box retailers the catalog doesn't cover. New finds slide into the same grid; those links open Google Shopping's product page.

3. 3

   Dupe rewriteonly when you ask for a dupe

   Search "lululemon align dupe" and a small, fast language model (via OpenRouter) rewrites it into a brand-free description, then we keep only alternatives that cost meaningfully less than the original.

4. 4

   Cheaper and lookalike matcheswhen you tap a product or drop a photo

   "Find cheaper" and "More like this" ask the catalog for visually similar products — in the lower price tier of the same category, or at any price. A photo works the same way and is never stored.

Powered by

Shopify Global Catalog via UCP

Live product listings from thousands of stores, plus the similarity search behind "find cheaper", "more like this", and photo search.

Used: Every search, the homepage discovery wall, and every similar-item request.

Google Shopping via Apify

A second wave that covers big-box retailers. Each run is metered and capped daily.

Used: After the instant results on /find.

OpenRouter

Routes a small language model that turns "dupe" requests into brand-free queries.

Used: Only for dupe and alternative searches.

Parallel

Web search and task research that powered nested's original deep hunts — the saved, shareable comparison pages at /s/….

Used: Deep hunts only. New searches start on the instant path above.

Exa

Web search and page contents used by deep hunts to find listings and product photos on retailer pages.

Used: Deep hunts only.

Product photos load straight from each store's image server. We don't copy or keep them.

For AI agents

Agents are welcome. Start with /llms.txt for the short version, or /llms-full.txt for parameters and example responses. Everything below is read-only, needs no key, and returns JSON.

JSON endpoints

• GET /api/live-search?q=…[&max=]

  Instant catalog results for a query (max = whole-dollar ceiling).

• GET /api/live-search?q=…&source=google

  The slower big-store wave for the same query.

• GET /api/similar?id=gid://shopify/p/…&mode=cheaper|similar

  Cheaper or lookalike products for a catalog item.

• GET /api/link-preview?url=…

  Reads a public product page and returns its name, price, and a search query.

• GET /api/discover?dept=…

  The homepage discovery wall for a department.

WebMCP tools

In browsers that support WebMCP, every nested page registers these tools, so an in-browser agent can call them instead of clicking around. None of them buy anything.

• search_productsAnswer

  Search and return title, price, store, URL, and product id.

• find_cheaperAnswer

  Cheaper lookalikes for a product id.

• find_similarAnswer

  Same style at any price for a product id.

• open_resultsAction

  Opens the /find results page for a query in this tab.

Honest limits

• Prices, stock, shipping, and returns belong to the store. Always confirm on the store page before you buy.
• The catalog is strongest for independent and direct-to-consumer brands. Big-box coverage comes from the second wave and can be thinner.
• Results are live and never cached, so the same search can differ minute to minute. That is also why we don't store them.
• nested compares and links out. It doesn't check out, hold your card, or buy anything for you — and neither do its agent tools.
• Searches are rate-limited per visitor, and the big-store wave has a daily budget. When it's paused you'll still get catalog results.