Apify MCP Server — 8,000+ Web Scrapers for Agents

Main

• Prefer the hosted endpoint when you want OAuth and the newest features (README calls out structured output schema inference not available in local stdio mode).

• Use discovery tools first (search-actors, fetch-actor-details) to pick the right Actor and only then run a paid/scraping-heavy call.

• If you want agents to pay without an API token, README describes x402 and Skyfire flows; start in a sandbox and understand billing before production use.

Source-backed notes

• README recommends connecting to the hosted server at https://mcp.apify.com for the best experience and latest features.
• README says the server turns Apify Actors into dynamically discoverable MCP tools and lists example Actors like Google Search and RAG Web Browser.
• README documents agentic payments (x402 and Skyfire) and notes mcpc supports x402 (npm install -g @apify/mcpc).

FAQ

• Is the hosted server required?: No — README describes local runs, but recommends hosted for latest features and OAuth.
• Do I need an Apify API token?: Not always — README describes x402/Skyfire payment modes; OAuth/token are other options.
• How do I keep scraping costs predictable?: Use discovery first, cap Actor inputs, and treat paid runs as an explicit step in your agent prompt.