Apify MCP Server

What apify-mcp-server does

apify-mcp-server gives MCP clients access to Apify Store Actors and Apify account operations. Actors are ready-to-run tools for activities such as fetching web pages, monitoring search results, generating leads, enriching data, and collecting information from e-commerce or social platforms. The agent can search the Store, inspect an Actor before using it, run the Actor, and retrieve its output.

The server also supports Actors that expose their own MCP tools and Server Actors with HTTP APIs. Beyond execution, it covers run inspection, datasets, key-value stores, saved Actor tasks, schedules, and selected Actor administration. It can search and retrieve Apify and Crawlee documentation as well.

How it works

A typical workflow starts with search-actors, followed by fetch-actor-details to obtain the Actor's input schema, README content, pricing, and, for Server Actors, documented HTTP operations. The client then uses call-actor with the appropriate input. That call returns run status and storage identifiers, which can be used with dataset or key-value-store tools to read the results.

For an Actor that runs an MCP server, fetch-actor-mcp-tool-list exposes its available tools and schemas before call-actor-mcp-tool invokes one. Server Actors can instead receive an HTTP request through send-server-actor-request. Run logs, status, cancellation, datasets, and stored records are available through separate tools.

Setup and configuration

The hosted endpoint is https://mcp.apify.com. Configure it in an MCP client that supports remote servers and authenticate with Apify through OAuth. If the client cannot use OAuth, send an Apify API token as an Authorization: Bearer <APIFY_TOKEN> header. Keep the token out of source repositories.

For local use, install or run the @apify/actors-mcp-server npm package through npx. Local stdio execution requires Node.js 22 or later and uses the APIFY_TOKEN environment variable. A token saved through apify login can also be read by the local server. The Apify MCP configurator generates client-specific configuration, and the Apify CLI can install configuration for supported clients such as Cursor, Claude Code, VS Code, and Codex.

Tools and capabilities

• Search Apify Store Actors and inspect their schemas, pricing, documentation, and output schema when requested.
• Run Actors, check run status, read logs, list runs, and abort active runs.
• Discover and invoke tools exposed by Actors that run MCP servers.
• Send documented HTTP requests to Server Actors.
• Read dataset metadata and items with filtering and pagination, and generate a JSON schema from dataset contents.
• Inspect key-value stores, list their keys, and retrieve stored values.
• List, create, update, publish, and unpublish saved Actor tasks.
• Create and inspect schedules that run Actors or tasks on cron expressions.
• Search and fetch Apify and Crawlee documentation.
• List or delete Actors in the connected account, and report problems to the Apify team.

Limitations and notes

Actor execution and result handling depend on the selected Actor's input schema, storage output, and API behavior. Fetch the Actor details before calling it so the client can construct valid input. Deleting an Actor is permanent, and public Actors must be unpublished before deletion. Authentication is required for the hosted or local connection, using OAuth or an Apify API token depending on the client setup.