Give your AI agent real product data from any store

Connect your agent over MCP or the API and give it a multi-store product catalog: full catalogs and the same product matched across stores, with prices and stock it can cite.

An agent that compares stores needs the data before the question.

Ask an AI assistant where a product is cheapest and it guesses, or tries to browse. Many stores block automated visitors, and one page read once is not a price list. To answer across stores, the agent needs every store's catalog, the same product matched across them, and prices and stock with the time they were read.

How it works

1
ConnectAdd DeepScout to your agent.One command in Claude Code or Codex, a custom connector in Claude, or an API key for your own agent.
2
Collect · you approveThe agent gathers the data.It adds the stores, shows you each quote and waits for your yes before it scrapes or matches.
3
AnswerIt answers from real data.Prices, stock and links from the stores' own pages, the same product matched across them.

Try it now

Sign in and create an API key for your own agent. Claude Code, Codex and Claude connect over MCP and sign you in themselves.

Open the console

What the agent works with

One product from our UK hi-fi run, matched across three stores and joined with each store's product file. This is what the agent reads before it answers.

What the agent can do with it

Two stores ask £199, but only one has it in stock. Price alone would pick the wrong store.

matches joined with the product files

{
  "product": "KEF Ci160CRDS in-ceiling speaker",
  "offers": [
    {
      "website_id": "richersounds.com-gb-gbp",
      "name": "KEF Ci160CRDS",
      "price": 279,
      "currency": "GBP",
      "availability": "Pre-order now",
      "availability_extracted": "ON_DEMAND",
      "product_url": "https://www.richersounds.com/kef-ci160crds/",
      "last_seen": 1793948365428
    },
    {
      "website_id": "petertyson.co.uk",
      "name": "KEF Ci160CRDS Round In-Ceiling Speaker",
      "price": 199,
      "currency": "GBP",
      "availability": "In Stock.",
      "availability_extracted": "IN_STOCK",
      "product_url": "https://petertyson.co.uk/kef-ci160crds-ceiling-speaker",
      "last_seen": 1793920474199
    },
    {
      "website_id": "sevenoakssoundandvision.co.uk",
      "name": "KEF Ci160CRDS In-Ceiling Speaker (Single)",
      "price": 199,
      "currency": "GBP",
      "availability": "Out of Stock Online",
      "availability_extracted": "OUT_OF_STOCK",
      "product_url": "https://www.sevenoakssoundandvision.co.uk/p-104896-kef-ci160crds-in-ceiling-speaker-single.aspx",
      "last_seen": 1793920554487
    }
  ]
}

An example answer

Built from the record above, the way an agent would answer it:

You: Where is the KEF Ci160CRDS cheapest, and can I get it now?

Agent: Peter Tyson and Sevenoaks both list it at £199, but only Peter Tyson has it in stock. Richer Sounds asks £279 and takes pre-orders.

What it costs

The data behind the example: 3 stores, 700 product pages each

Analyze3 stores€0

Scrape1,948 products€1.69

Match2,206 products × €7 per 1k€15.44

Total€17.13

Connecting over MCP costs nothing; you pay for the jobs, at the same prices as the API. Each job shows its quote first, and the charge never goes above 1.5× that quote.

Scrape your first store

Paste a store URL. You see the product count, a sample and the exact price before you pay. The analysis is free for 10 stores a month.

Questions

Which agents work?

Claude Code, Codex, Claude (web and desktop), Cursor, VS Code, and any client that supports remote MCP servers. For your own agent, use the REST API.

Can the agent spend money without asking?

Only approving a job charges credit. The server tells the agent to show you the price and wait for your answer first, so keep your client's confirmation prompt on for that tool. The agent can't top up credit.

Does the agent get live prices?

No. Jobs take minutes to hours, and the agent answers from the last run. Every product carries the time it was read, and you refresh with a scrape of the product URLs you care about.

Do the stores need barcodes?

No. Matching compares names, specs, prices and images, then checks each pair with AI.

Which stores work?

Online stores on any platform. Marketplaces like Amazon, eBay and Allegro are not supported.