ScrapingBot

What ScrapingBot MCP server does

The ScrapingBot MCP server exposes public web data as tools for MCP-compatible AI clients. It can retrieve website content as markdown or HTML, optionally render JavaScript, capture screenshots, and extract requested fields into JSON with an AI query or schema. Browser scenarios support actions such as clicking, filling forms, scrolling, and waiting.

The service also provides data access for Google, TikTok, Instagram, and Amazon. Available operations include web and vertical Google searches, reviews, public social profiles and media, comments and follower data, Amazon product and seller lookups, rankings, suggestions, and related searches. This places the server in the Search & Data Extraction category.

How it works

The hosted endpoint is https://scrapingbot.io/api/mcp. It uses stateless Streamable HTTP, so there is no local server package to install. Send the API key with the x-api-key header or an Authorization: Bearer header. Clients that accept only a URL can include the key as the api_key query parameter, which should be treated as private credentials.

Tool calls use the same credit, concurrency, timeout, and refund rules as the underlying REST APIs. Website scraping costs one credit for a plain request, five for rendered pages, and an additional five for AI extraction. Google and Amazon operations generally cost 10 credits, Instagram operations cost five, and TikTok operations cost one. Utility job and capability tools are free where identified in the API table.

Setup and configuration

Create a ScrapingBot account and obtain an API key before connecting a client. Claude Code can register the endpoint with its HTTP transport and an x-api-key header. Claude Desktop can use a custom connector URL; because that connector accepts a URL, the key is appended as ?api_key=.... Cursor and VS Code use JSON MCP configuration with the hosted URL and an x-api-key header.

For direct API use or Claude Code Agent Skills, set SCRAPINGBOT_API_KEY in the environment where the agent runs. The repository also includes examples in curl, Python, and Node, plus skills covering web scraping, Google, TikTok, Instagram, and Amazon tasks.

Tools and capabilities

The ScrapingBot MCP server lists 23 tools across these areas:

• Website access: scrapeWebsite, extractStructuredData, and runBrowserScenario.
• Google: googleSearch and googleReviews.
• Instagram: instagramUser, instagramSearch, instagramMedia, and instagramFollowers.
• TikTok: tiktokVideo, tiktokUser, tiktokSearch, tiktokComments, and tiktokFollowers.
• Amazon: amazonSearch, amazonProduct, amazonProducts, amazonSuggestions, and amazonRankings.
• Utilities: listCapabilities, getScrapeJob, pollJobUntilDone, and providerRequest.

providerRequest can call supported underlying API endpoints, while job tools can retrieve or poll scrape results. Scrape jobs remain available for 24 hours.

Limitations and notes

Requests are limited by concurrent in-flight work rather than a per-minute quota. The free plan allows one concurrent request; paid plans allow between 10 and 200. Exceeding the limit returns HTTP 429. Timeouts are 45 seconds for scraping and Instagram, 30 seconds for TikTok and Amazon, and 15 seconds for Google.

The service is intended for public data. API calls require credits, and the account must have enough balance for the selected operation. Errors such as invalid parameters, authentication failures, timeouts, concurrency limits, and most upstream failures are not charged. Website responses that represent a successful page response can still incur the website request charge, including a genuine target-site 400 or 404.