Skip to main content
Glama
amazon-scraper-api

amazon-scraper-api

amazon-scraper-api-mcp

npm npm downloads license Glama MCP server

MCP (Model Context Protocol) server for Amazon Scraper API. Plugs into Claude Desktop, Cursor, Claude Code, Continue, or any MCP-compatible AI client. Gives your model live Amazon product data as a first-class tool call.

What it unlocks

"Find me the highest-rated wireless earbuds under $150 on amazon.com, then check if they're cheaper on amazon.de"

That's one prompt. Without MCP, your AI can't fetch Amazon pages (Amazon blocks LLM browsing) and has zero recency for prices and stock. With this MCP server, it calls amazon_search + amazon_product directly and comes back with structured data from Amazon Scraper API.

Related MCP server: Amazon Price Tracker MCP

Tools exposed

Tool

What it does

Typical use

amazon_product

Fetch one product by ASIN or URL

"get price + rating for B09HN3Q81F"

amazon_search

Keyword search with sort and pagination (one page per call)

"top 10 cast iron skillets under $50"

amazon_batch_create

Queue up to 1000 ASINs for async scrape

"scrape all 500 products in my catalog, webhook me when done"

amazon_batch_status

Poll a batch's progress

"how much of batch xyz is done?"

Each returns structured JSON: title, price, rating, reviews count, availability, buybox, variants, images, bullets, categories, spec tables.

Benchmark (live production, 2026-04)

Metric

Value

Median latency (product, US)

~2.6 s

P95 latency

~6 s

Price / 1,000 requests

$0.50 flat

Marketplaces

20+

Claude Desktop setup

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "amazon-scraper": {
      "command": "npx",
      "args": ["-y", "amazon-scraper-api-mcp"],
      "env": {
        "ASA_API_KEY": "asa_live_..."
      }
    }
  }
}

Restart Claude Desktop. Tools appear under the MCP icon in the chat composer.

Cursor setup

Settings → MCP → Add server:

  • Command: npx -y amazon-scraper-api-mcp

  • Env: ASA_API_KEY=asa_live_...

Claude Code setup

claude mcp add amazon-scraper -- npx -y amazon-scraper-api-mcp
# then set ASA_API_KEY in the environment Claude Code runs in

Example conversation (Claude Desktop)

You: What's the current price of AirPods Pro 3rd gen on amazon.com?

Claude: [calls amazon_product with ASIN B09HN3Q81F]

The AirPods Pro 3rd gen are currently $199.00 on amazon.com, down from $249.00 (20% off). They have a 4.7-star rating from 58,214 reviews and are in stock, shipping from and sold by Amazon.com with Prime.

You: Compare that to the German Amazon listing.

Claude: [calls amazon_product with query=B09HN3Q81F, domain=de]

On amazon.de the same product is listed at 229.00 EUR. At today's exchange rate that's about $245, roughly 23% more than the US price. German listing ships from Amazon and qualifies for Prime delivery.

Why this vs. a generic "browse the web" MCP

Generic browser tools that try to load amazon.com typically get blocked (robot check) or serve a mobile-stripped page. This server routes every call through Amazon Scraper API, which:

  • Detects and retries robot/CAPTCHA pages through escalating proxy tiers

  • Returns structured JSON (typed fields), not HTML soup

  • Supports 20+ marketplaces with country-matched residential IPs

  • Handles batch (hundreds to thousands of ASINs) with webhook delivery

Error handling

Errors are surfaced to the model as tool errors whose message carries the HTTP status and the API's error code, for example HTTP 502: target_unreachable or HTTP 402: INSUFFICIENT_CREDITS. The model then decides whether to retry or abandon the sub-task. The server itself does not retry.

Common codes: INVALID_API_KEY, INSUFFICIENT_CREDITS, RATE_LIMITED, CONCURRENCY_LIMIT, product_not_found, target_unreachable (retry after 60s or more), extraction_failed, SERVICE_OVERLOADED. Full table: amazonscraperapi.com/docs/guides/errors.

Get an API key

app.amazonscraperapi.com. 1,000 free requests on signup, no credit card required. Enough to test every tool this MCP exposes plus a few dozen productive chats.

License

MIT

Available Tools

4 tools
amazon_batch_createA

Queue up to 1000 ASINs or search queries for async processing. Returns a batch id - poll with amazon_batch_status or receive a webhook callback.

Inputs: endpoint (required), items (required), webhook_url.

amazon_batch_statusB

Poll an async batch job for progress + results.

Inputs: id (required).

amazon_productB

Fetch structured data for a single Amazon product by ASIN. Returns ~55 fields including title, price, variations, reviews, category ladder, images.

Inputs: query (required), domain, language.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.8/5.0

    Scored across 4 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: batch creation, batch status polling, single product fetch, and keyword search. No overlap between tools.

    Naming Consistency4/5

    All tools use snake_case and start with 'amazon_'. Three follow verb_noun pattern ('batch_create', 'batch_status', 'search'), while 'amazon_product' is a noun implying fetch. Minor inconsistency but overall predictable.

    Tool Count5/5

    Four tools cover the essential scraping operations: search, single product details, and async batch processing. The count is well-scoped for the server's purpose.

    Completeness4/5

    Covers search, single product, and batch processing. Missing dedicated review or category tools, but the product tool includes reviews. Minor gaps but core workflows are complete.

    Maintenance

    ActivityMaintained
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers