@apify/actors-mcp-server

What it does

The Apify MCP Server connects Claude and other MCP clients to thousands of ready-made web scrapers and automation tools from the Apify Store. It enables agents to extract structured data from social media platforms (Facebook, Instagram), search engines (Google), maps, e-commerce sites, and arbitrary websites without writing scraper code. The server can be deployed as a hosted HTTPS endpoint at mcp.apify.com (OAuth-compatible) or locally via stdio with an API token.

Who it's for

Backend engineers and data engineers building AI agents that need web scraping, structured data extraction, or automated web interaction as part of their pipeline. Also suitable for researchers and analysts who want Claude to fetch fresh data from the web without managing custom scraping infrastructure.

Common use cases

• Extract contact details (emails, phone numbers) from Google Maps for a given location or business query.
• Scrape social media feeds (Facebook posts, Instagram profiles) to feed into downstream analysis or reporting workflows.
• Query Google Search Engine Results Pages (SERPs) and retrieve ranked links with snippets for a given search term.
• Crawl arbitrary websites and retrieve their content in structured form for RAG or document processing.
• Automate web interactions (form submission, login) on e-commerce or target sites without writing custom code.

Setup pitfalls

• APIFY_TOKEN and secrets management — the codebase contains multiple secret patterns; inject the token via environment variables, never commit to version control.
• OAuth vs. token auth — the hosted mcp.apify.com endpoint supports OAuth (recommended for Claude.ai); local stdio mode requires explicit APIFY_TOKEN.
• SSE transport deprecation — SSE support ends April 1, 2026; update MCP client configurations to use the newer streamable HTTP transport.
• High-risk runtime — this server makes arbitrary network calls and reads/writes the filesystem; sandbox it if invoking untrusted Apify Actors.