Apify MCP Server

What is Apify MCP Server?

Apify MCP is Apify's official Model Context Protocol server: it lets AI agents search Apify Store and run thousands of ready-made scrapers, crawlers and automation tools (Actors) to pull data from social media, search engines, maps, e-commerce sites and any other website.

The server is open source (MIT) and maintained by Apify in the apify/apify-mcp-server repository (the old apify/actors-mcp-server URL redirects there). Connect to the hosted endpoint https://mcp.apify.com (Streamable HTTP with OAuth) or run the @apify/actors-mcp-server npm package locally over stdio. Apify recommends the hosted server, which also adds output schema inference that the local version lacks.

How to use Apify MCP Server?

You need an MCP client and usually an Apify account. An API token (from API & Integrations in Apify Console) is only needed for local stdio or the Bearer token option; the hosted server uses OAuth. The MCP configurator generates configs for most clients.

Hosted server with OAuth (recommended):

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com"
    }
  }
}

Hosted server with a Bearer token instead of OAuth:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com",
      "headers": {
        "Authorization": "Bearer <APIFY_TOKEN>"
      }
    }
  }
}

Local stdio, for clients that do not support remote servers (requires Node.js):

{
  "mcpServers": {
    "actors-mcp-server": {
      "command": "npx",
      "args": ["-y", "@apify/actors-mcp-server"],
      "env": {
        "APIFY_TOKEN": "YOUR_APIFY_TOKEN"
      }
    }
  }
}

With the Apify CLI, apify mcp install cursor (or another client name) writes the config for you.

Claude Code

Apify ships a Claude Code plugin that bundles the MCP server with Agent Skills. Inside Claude Code run:

/plugin marketplace add apify/apify-claude-code-plugin
/plugin install apify@apify
/reload-plugins

Then run /mcp, open plugin:apify:apify, select Enable, then Authenticate and sign in. In headless environments, set APIFY_TOKEN before starting Claude Code.

Claude Desktop

Open Settings > Connectors, select Add > Add custom connector, enter a name such as Apify and the URL https://mcp.apify.com, keep the detected settings and select Add. It also works on claude.ai. Alternatively, install Apify from the connector directory.

Cursor

In the Cursor agent chat, run /add-plugin apify (or install Apify from the Cursor Marketplace). Then open Cursor Settings > Tools & MCPs, select Connect on the Apify MCP server and sign in.

VS Code

Open the Extensions view, search @agentPlugins apify, install the apify entry from the Awesome Copilot marketplace and confirm the trust prompt. When VS Code asks to authenticate the MCP server, select Allow and sign in to Apify.

Codex

In the Codex CLI, run /plugins, open Add Marketplace, enter apify/apify-codex-plugin, then install Apify from the Apify Plugin tab. Sign-in opens on the first call that needs authentication.

Key features of Apify MCP Server

• Turns any Apify Actor into an MCP tool by loading the Actor's input schema, so the model knows which arguments to pass.
• Dynamic discovery: the agent can find a suitable Actor with search-actors and run it with call-actor without pre-configuring it.
• Pre-configured with the apify/rag-web-browser and apify/web-fetch Actors plus helper tools.
• Configurable tool set through the tools URL parameter or --tools CLI flag (categories, single tools or specific Actors).
• OAuth or Bearer token authentication; a few discovery and docs tools work without any account.
• Agentic payments through AGI, direct x402 (USDC on Base) or Skyfire, so agents can pay for runs without an API token.
• Optional UI mode (ui=true) for interactive MCP App widgets; telemetry can be switched off.

Apify MCP Server tools

Tools are grouped into selectable categories:

• actors (default): search-actors, fetch-actor-details and call-actor find an Actor, read its input schema, pricing and README, then run it.
• docs (default): search-apify-docs and fetch-apify-docs pull Apify documentation into context.
• Actor tools (default): apify--rag-web-browser searches the web and returns content from top results; apify--web-fetch fetches a URL as Markdown, text, HTML or links.
• runs and storage: get-actor-run, get-actor-run-log, abort-actor-run, get-dataset-items, get-key-value-store-record and related tools. When an Actor tool or call-actor is loaded, the server adds get-actor-run, get-dataset-items, get-key-value-store-record and abort-actor-run automatically.
• tasks, schedules, builds and api: manage saved tasks, cron schedules and builds, or query the Apify API with apify-api-search and the read-only apify-api-read.

Note that call-actor returns the run status and storage IDs, not the scraped data; the agent then reads results with get-dataset-items.

Use cases of Apify MCP Server

• Web research: "Find the latest news about autonomous AI agents and summarize them."
• Local lead lists: "Scrape the top 10 restaurants in Prague from Google Maps with their contact details."
• SEO research with the Google Search Results Scraper: "Get the first page of Google results for 'apify mcp' and list the ranking domains."
• Social monitoring: extract posts and comments from several Facebook pages, or Instagram posts and profiles, for analysis.
• Paging results: "Show me the next 10 results from that run."
• Automation: create an Actor task and a schedule so a scraper runs on a cron cadence.

Apify MCP Server vs alternatives

Firecrawl MCP Server offers scraping, crawling and search through one API, while Apify gives the agent a marketplace of site-specific Actors (Google Maps, Instagram, Facebook and more). Playwright MCP Server and Puppeteer MCP Server drive a local browser step by step; Apify runs the work in its cloud and returns structured datasets. This listing is the official Apify server.

Troubleshooting Apify MCP Server

• Unauthorized or invalid token: check the token in Apify Console under API & Integrations, make sure APIFY_TOKEN is set for local use, or remove and re-add the connector to rerun OAuth.
• Old /sse URL stops working: the legacy SSE endpoint has been removed. Use https://mcp.apify.com without the /sse suffix.
• Tools missing in Claude Desktop: check that Apify tools are not blocked in connector settings, restart the app, or remove and re-add the connector.
• Local server will not start: confirm Node.js is installed (node -v), use @apify/actors-mcp-server@latest, and clear a corrupted npx cache with rm -rf ~/.npm/_npx on macOS or Linux.
• Slow or empty results: check the run log in Apify Console, and fetch items with get-dataset-items, since call-actor does not return them.