Apify MCP Server 🔌 MCP server Freemium

What it does

Apify's official MCP server lets an agent use Actors from Apify Store as tools. Actors are ready-made scrapers, crawlers and automation jobs for search engines, maps, social networks, e-commerce sites and ordinary web pages. The agent can search the store, read an Actor's input schema and pricing, run it, and then page through the results. For coding work, the default web browsing and web fetch Actors can pull a page or search results into context as Markdown. The same tools also help when you are building or debugging your own Apify scrapers, because the agent can inspect runs, logs and stored data. The server itself is MIT-licensed and maintained by Apify.

Tools it gives your agent

• Actors: search-actors, fetch-actor-details and call-actor, plus any Actor loaded as its own tool
• Web: the apify--rag-web-browser and apify--web-fetch Actor tools, enabled by default
• Runs and storage: get run details and logs, abort runs, and read datasets and key-value stores
• Docs: search and fetch Apify documentation
• Optional: manage saved Actor tasks and cron schedules

Works with

The recommended option is the hosted remote server at https://mcp.apify.com, which uses Streamable HTTP with OAuth or a bearer token. Some newer features only work on the hosted version. A local stdio version runs through npx. The README lists Claude Code, Claude Desktop, Cursor, VS Code, OpenCode and Kiro, and any MCP client should work.

How to connect

Point a remote-capable client at the hosted URL, or run it locally with an APIFY_TOKEN environment variable:

https://mcp.apify.com
npx @apify/actors-mcp-server

Cost and limits

The server is free, but running Actors uses your Apify account's usage and each Actor's own pricing. Looking up Actors and docs is free. Agents can also pay without an account through x402, Skyfire or Apify's prepaid AGI tokens.

Safety notes

The token gives the agent access to your Apify account, and the optional schedule and task tools can change it. Use the tools parameter to load only what you need. Telemetry is on by default, and you can turn it off with --telemetry-enabled=false. Scraping third-party sites is your responsibility, so check their terms.