scrape-do-mcp

中文文档 | English

An MCP server that wraps Scrape.do's documented APIs in one package: the main scraping API, Google Search API, Amazon Scraper API, Async API, and a Proxy Mode configuration helper.

Official docs: https://scrape.do/documentation/

Coverage

• `scrape_url`: Main Scrape.do API with JS rendering, geo-targeting, session persistence, screenshots, ReturnJSON, browser interactions, cookies, and header forwarding.
• `google_search`: Structured Google SERP API with `google_domain`, `location`, `uule`, `lr`, `cr`, `safe`, `nfpr`, `filter`, pagination, and optional raw HTML.
• `amazon_product`: Amazon PDP endpoint.
• `amazon_offer_listing`: Amazon offer listing endpoint.
• `amazon_search`: Amazon search/category endpoint.
• `amazon_raw_html`: Raw HTML Amazon endpoint with geo-targeting.
• `async_create_job`, `async_get_job`, `async_get_task`, `async_list_jobs`, `async_cancel_job`, `async_get_account`: Async API coverage with both MCP-friendly aliases and official field names.
• `proxy_mode_config`: Builds official Proxy Mode connection details, default parameter strings, and CA certificate references.

Compatibility Notes

• `scrape_url` supports both MCP-friendly aliases and official parameter names:
  • `render_js` or `render`
  • `super_proxy` or `super`
  • `screenshot` or `screenShot`
• `google_search` supports:
  • `query` or `q`
  • `country` or `gl`
  • `language` or `hl`
  • `domain` or `google_domain`
  • `includeHtml` or `include_html`
• `async_create_job` accepts both alias fields like `targets`, `render`, `webhookUrl` and official Async API fields like `Targets`, `Render`, `WebhookURL`.
• `async_get_job`, `async_get_task`, and `async_cancel_job` accept both `jobId`/`taskId` and official `jobID`/`taskID`.
• `async_list_jobs` accepts both `pageSize` and official `page_size`.
• For header forwarding in `scrape_url`, pass `headers` plus `header_mode` (`custom`, `extra`, or `forward`).
• Screenshot responses preserve the official Scrape.do JSON body and also attach MCP image content when screenshots are present.
• `scrape_url` now defaults to `output="raw"` to match the official API more closely.
• `scrape_url` includes response metadata in `structuredContent`, which helps surface `pureCookies`, `transparentResponse`, and binary responses inside MCP.

Installation

Quick Install

claude mcp add-json scrape-do --scope user '{
  "type": "stdio",
  "command": "npx",
  "args": ["-y", "scrape-do-mcp"],
  "env": {
    "SCRAPE_DO_TOKEN": "YOUR_TOKEN_HERE"
  }
}'

Claude Desktop

Add this to `~/.claude.json`:

{
  "mcpServers": {
    "scrape-do": {
      "command": "npx",
      "args": ["-y", "scrape-do-mcp"],
      "env": {
        "SCRAPE_DO_TOKEN": "YOUR_TOKEN_HERE"
      }
    }
  }
}

Get your token at https://app.scrape.do

Available Tools

Example Prompts

Scrape https://example.com with render=true and wait for #app.

Search Google for "open source MCP servers" with google_domain=google.co.uk and lr=lang_en.

Get the Amazon PDP for ASIN B0C7BKZ883 in the US with zipcode 10001.

Create an async job for these 20 URLs and give me the job ID.

Development

npm install
npm run build
npm run dev

License

MIT