Amazon data for product and market research

amazon entry points. REST and MCP are two interfaces to the same tools.

The MCP mount point has the same shape as other live platformssame shape: the same host and /mcp/ prefix, with only the final segment set to amazon. Replace the final segment with another platform name to get its endpoint.

Choose any of the three setup methods. Authentication, metering, and caching stay in the cloud; no upstream keys are stored locally.

# Claude Code: connect with one command
claude mcp add --transport http claudebaby-amazon https://claudebaby.cn/mcp/amazon \
  --header "Authorization: Bearer $CB_KEY"

# After connecting, use /mcp in a session to see the tools

{
  "mcpServers": {
    "claudebaby-amazon": {
      "type": "http",
      "url": "https://claudebaby.cn/mcp/amazon",
      "headers": { "Authorization": "Bearer <your API key>" }
    }
  }
}

# Local stdio: no secrets are stored locally; data still comes through the cloud gateway
uv tool install --from "git+https://github.com/lingling1989r/claudebaby.git#subdirectory=mcps" claudebaby-mcp
CLAUDEBABY_API_KEY=cb_live_xxx claudebaby-mcp --platform amazon

Issue keys in the console; plaintext is shown once. Insufficient balance returns insufficient_balance（HTTP 402）。

The tool name is the final REST path segment; the same tools are available through MCP and return the same envelope.

Total 20 endpoints.

The example uses anker as the search term and calls amazon’s category_name_search. The search term is placed in the tool’s actual input field; other required fields remain placeholders. Replace <…> with your value. See the endpoint catalog above for parameters and descriptions, or https://claudebaby.cn/v1/ecommerce/amazon/tools.

export CB_KEY=cb_live_xxx

curl -s -X POST https://claudebaby.cn/v1/ecommerce/amazon/tools/category_name_search \
  -H "Authorization: Bearer $CB_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name": "anker"}'

{
  "ok": true,
  "data": {
    "…": "category_name_search response; fields vary by tool"
  },
  "meta": {
    "platform": "amazon",
    "domain": "ecommerce",
    "tool": "category_name_search",
    "request_id": "req_…",
    "credits": "<credits used by this call; see /pricing.json>"
  }
}

Run the request to see a real response (endpoint and route billing applies) — the example above data above is only the envelope, not fabricated sample data. To try it in a browser, use API details and enter your own key.

Choose top-ups or a plan; credits are reserved before a paid fetch and insufficient balances are rejected.

Standard e-commerce costs 1 credit per execution. A ×N badge means N credits per execution. Composite subcalls and pages are billed separately. Credits are held before execution, settled on success and released on failure; timeouts await reconciliation.

Unpriced endpoints are temporarily unavailable. Fallback quotes include the cost budget for the failed primary attempt. See the actual invocation ledger.

Charge per call = credits per execution × executions; 1 credit = ¥0.02. See the tier for each endpoint in /pricing.json (including per-platform prices).