Amazon Products MCP server

How an agent would call it

Two steps: search returns refs, then data_fetch opens one of them. The arguments below are the arguments that were sent.

{
  "tool": "search",
  "arguments": {
    "query": "mechanical keyboard",
    "sources": [
      "amazon"
    ],
    "limit": 2,
    "runId": "<from run_begin>"
  }
}

{
  "tool": "data_fetch",
  "arguments": {
    "sourceId": "amazon",
    "ref": {
      "id": "B08Z6X4NK3"
    },
    "runId": "<from run_begin>"
  }
}

This runs against an account — send Authorization: Bearer <token> with the call. Get one at app.bedrocktap.com/dashboard/tokens; the free tier needs no card, and it reaches amazon like every other source.

What it costs. 6 credits per search call and 6 per fetch, whatever the row count. The search above returned 2 products for 6 credits; the fetch cost 6.

What came back

{
  "ref": {
    "id": "B08Z6X4NK3",
    "url": "https://www.amazon.com/dp/B08Z6X4NK3",
    "title": "Logitech G413 SE Full-Size Mechanical Gaming Keyboard - Black"
  },
  "text": "Upgrade your gameplay experience with the Logitech G413 SE Full-Size Mechanical Gaming Keyboard. It delivers a gaming-grade mechanical switch experience. It’s designed to help serious players perform at a higher level and push themselves even further.…",
  "fetchedAt": "2026-09-12T15:03:59.741Z"
}

Each ref carried price, originalPrice, currency, rating, ratings, isPrime and country in its meta — $69.99 down from $89.99, 4.5 stars across 1,654 ratings for the first hit; $69.99 and 4.6 across 1,794 for the second. The fetched payload widened to the whole listing: product_description, about_product, product_details, category_path, rating_distribution, sales_volume, product_num_offers, product_variations, delivery, top_reviews and the photo and video lists.

Observed 2026-09-12. One call against a live source — ask again tomorrow and the rows move.

What a row contains

One row is one product: its ASIN as the id, the product page, and the listing title. meta carries the current price, the struck-through original price, the currency, the star rating, the number of ratings, whether it ships on Prime, and the marketplace the row came from. The marketplace travels on the ref so a fetch reads the same country the search did. Rating counts are a demand proxy; units sold and seller revenue are not published anywhere here.

A question it answers. What does this category actually sell for, and how many ratings does the leader have?

Routing guidance

What an agent reads to decide whether amazon is the right source, and what to reach for when it is not.

Answers

Retail pricing, competitive product research and category structure. Search results, product detail, offers, best sellers, deals and seller info.

Query shape

Plain product keywords. Marketplace and depth are set per call.

Price

Priced in credits, per row, per query or per call depending on the source. Every call is quoted before it runs — `cost_estimate` returns the exact figure and spends nothing. See /pricing.

Latency

Fast (~1-3s).

Prefer instead

`vinted` for second-hand European resale and sell-through data. `google-serp` for where a product ranks in search rather than on the storefront.

Cannot answer

Deep review text — reviews here are shallow; use `app-store-reviews` for apps and `trustpilot` for brands. Search volume. Anything off Amazon.

Related sources