MCP
OAuth, no key to paste. Works with Claude, Cursor, and any MCP client.
claude mcp add --transport http storerouter https://mcp.storerouter.ai/mcpEcommerce data for AI agents
One API forall stores.
Search across stores, parse any product page, and find alternatives — in one enriched schema agents can comprehend without scraping.
RESTMCPCLISDKsAgentic web search finds pages.
StoreRouter finds products.
Web search APIs return links and snippets. Your agent still has to open each page, pull out the product, and check whether it actually matches. StoreRouter searches the catalogs themselves, so every result is a product you can filter, sort and compare.
woody eau de parfum under $150
Agentic web search
Editorial · listicle
The 15 best woody perfumes, according to our editors
…from cult niche picks to everyday signatures. Prices at time of publishing…
Store · category page
Eau de Parfum | Women's & Men's Fragrance
Shop 500+ scents. Free samples with every order…
Forum · thread
Woody perfume recommendations under $150?
…looking for something smoky but not too sweet. Open to niche…
Pages. Your agent still opens each one.
Covers the whole web
StoreRouter
Comme des Garçons · doverstreetmarket.com
Wonderwood Eau de Parfum
$125.00
- 50 ml
- 100 ml
Notes: cedar · incense · oud · vetiver
Yves Saint Laurent · sephora.com
L'Homme Eau de Parfum
$138.00
- 100 ml
Notes: citrus · orange · vetiver · woody
Glossier · sephora.com
You Doux Eau de Parfum
$82.00
- 8 ml
- 50 ml
- 100 ml
Notes: frankincense · myrrh · smoke · violet
Products. Filter by price, size, notes and stock.
Products from the StoreRouter index, October 2026. Search hits are illustrative.
Crawlers fetch pages.
StoreRouter is a live catalog.
A crawler returns page markup, and leaves the rest to you: parse fields, wrangle per-site schema, schedule refreshes, enrichment and indexing. StoreRouter abstracts all of that away into a live catalog.
| Crawler APIs | StoreRouter | |
|---|---|---|
| Fetch | Any page | Crawled for you |
| Parse | A parser per site | Variants, per-size stock |
| Normalize | Map each site to your schema | One schema across stores |
| Enrich | Tag attributes yourself | Fit, skin type, scent notes |
| Refresh | Your re-crawl schedule | Re-crawled as stores change |
| Index + search | Your own index | Built in, across every store |
| Tokens per product | ~10.8K as page Markdown | ~1.3K in our tests |
Every store describes products its own way.
StoreRouter keeps the facts, and adds meaning.
Every product keeps the store's own facts, mapped into one schema, then gets the attributes the page only implies, for its category: fit for a dress, skin type for a moisturizer, notes for a perfume, upholstery for a sofa. More for your agent to search, filter and compare on, across every store.
See schema and facets for Reformation.
From the store
Everything the merchant publishes.
- title
- price & sale price
- variants
- per-size stock
- images
- product URL
Normalized
Made comparable across stores.
- canonical brand
- taxonomy path
- color family
- currency
Enriched by StoreRouter
What the page implies but never labels, by category.
+ more
- fit
- fabric
- neckline
- length
- pattern
- occasion
Ditch the browser. 100× faster, 500× cheaper.
To read one product, an agent usually drives a headless browser: launch a session, route through a proxy, wait out the bot check, render the JavaScript, then pay an LLM to read megabytes of HTML. That's seconds per product, a pile of tokens, and it breaks whenever a store blocks bots. And because a model does the reading, ask twice and you can get two different answers.
StoreRouter reads each store with code, not a model. One call returns the product, structured and enriched, in about 100 ms, and the same call returns the same answer every time.
- StoreRouter~100 ms
100× faster · 500× cheaper than a headless browser
- Fetch the HTML~5 s
- proxy
- fetch + retry
- LLM read
missing: pricemissing: variantsmissing: stockrendered by JavaScript, not in the HTML
~500K tokens · 3 of 7 stores blocked
- Headless browser~10 s
- launch browser
- proxy
- bot check
- render JS
- scroll
- LLM read
~800K tokens · often blocked, then retry
Same product, complete, in a fraction of the time and tokens — and it doesn't break when a store blocks bots.
Cost compares StoreRouter's $2 per 1,000 lookups with an LLM reading the page at ~$1.25 per million input tokens.
Stores redesign all the time.
That's our problem, not yours.
Scrapers break silently when a store changes. StoreRouter keeps answering the same way.
Thousands of stores, any of which can redesign at any time. When one does, like Zara here, StoreRouter re-learns that one store while the rest keep answering.
Plug it in
Wherever your agent runs.
Docs
CLI
Search, look up, and request coverage from the terminal.
npx -y @octogen-ai/cli@next search "velvet sofa"REST
Every endpoint, one Bearer key, an OpenAPI contract.
curl https://api.storerouter.ai/v1/products/search \
-H "Authorization: Bearer $STOREROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"q": "velvet sofa"}'SDKs
Typed TypeScript and Python clients.
npm install @octogen-ai/sdkPricing
Pay per successful request.
No subscriptions and no seat fees. Every organization gets free credit each month; top up when you need more.
Lookup & Parse
$2
per 1,000 requests
Search
$14
per 1,000 requests
Free credit
$5
every month
- Only successful 2xx responses are billed.
- Purchased credit never expires.
Give your agent every store.
Create an organization, mint a key, and make your first call in a couple of minutes.
Are you an AI agent? Fetch https://storerouter.ai/llms.txt to learn how to get product data from StoreRouter, and how to make a storefront's products readable by agents.