用 MCP 讓 AI 智能體取得各網店的即時商品價格

Ask an AI assistant "who sells the Sony WH-1000XM5 cheapest right now?" and you usually get one of two answers: a guess from training data, or a link to a single shop. Agents that compare products, monitor prices or research suppliers need something better: current listings from many shops, with the seller, the price and a link they can check.

That is what datawebot does. It is a hosted MCP server that gives agents ecommerce product data, so any MCP client (Claude, ChatGPT, Cursor, VS Code, Gemini CLI and others) can use it without you writing a scraper.

What your agent can do with it

• Search products across online shops. search_products takes a query and a country and returns titles, prices, shops, ratings, review counts, images and links.
• Compare every seller for one product. get_product_offers lists each seller's price, total with shipping, stock or delivery notes and a direct link to the product page.
• Pull structured data from store pages. Extraction jobs run on Amazon search and product pages in 10 countries, Walmart US and the other stores in the endpoint directory. The agent picks a contract with list_job_blueprints, submits a URL with submit_job, then reads records with get_results.

Results are dated snapshots with source links. There is no price history and no alerts, so treat prices as "as of" the retrieval time.

Connect it in one line

The quickest way is to let your agent set itself up. Paste this into Claude Code, Cursor, Codex or any agent that can read a URL:

Read https://mcp.datawebot.com/setup.md and follow the instructions.

The agent starts the sign-in, you approve it from your email inbox, and the server is added. Nobody pastes a key into the chat.

Or add the server yourself. The endpoint is Streamable HTTP with OAuth sign-in:

{
  "mcpServers": {
    "datawebot": {
      "type": "http",
      "url": "https://mcp.datawebot.com/mcp"
    }
  }
}

In Claude Code you can also install the plugin, which adds the server and a skill that teaches the agent how to use it:

/plugin marketplace add datawebot/datawebot-mcp
/plugin install datawebot@datawebot

For scripts, there is a REST API with an API key and an OpenAPI spec.

A typical agent flow

1. discover lists the tools available right now.
2. search_products with {"query": "Sony WH-1000XM5", "country": "US"} returns candidates.
3. get_product_offers with a result's product_id returns every seller and link.
4. For bulk or page-level data, list_job_blueprints → submit_job → get_job → get_results.

Pricing

Sign-in is free. Every account gets a free daily search allowance and 200 trial credits for extraction jobs. After that, credit packs start at $10, with no subscription, and failed or empty jobs are free.

Links

• Docs: https://www.datawebot.com/docs/agents
• MCP server: https://mcp.datawebot.com/mcp
• Official MCP Registry: com.datawebot/datawebot
• Connection kit (MIT): https://github.com/datawebot/datawebot-mcp