Actors MCP Server

What apify/actors-mcp-server MCP server does

The apify/actors-mcp-server MCP server connects an MCP-compatible AI client to Apify Actors. Actors are hosted tools in the Apify Store that can collect or process information from websites, social media, search engines, maps, e-commerce platforms, and other sources. The agent can search for suitable Actors, inspect their details, invoke them, and retrieve results from their datasets.

The repository describes examples including scrapers for Facebook posts, Google Maps contact details, Google search results, and Instagram content. It also references Actors for web browsing and URL fetching, including JavaScript rendering and anti-bot protection. The available result and behavior depend on the Actor selected and its input requirements.

How it works

Clients can connect to the hosted endpoint at https://mcp.apify.com using Streamable HTTP. The hosted service supports OAuth, so supported clients can connect with the URL, or requests can include an Authorization: Bearer header containing an Apify token. The README identifies the hosted endpoint as the recommended option and notes that it includes features such as output-schema inference for structured Actor results that are unavailable in the local stdio mode.

For local use, an MCP client starts the npm package as a standard input/output process. The process uses the APIFY_TOKEN environment variable to authenticate with Apify. Client support for dynamic tool discovery and related MCP features may vary.

Setup and configuration

The local command documented by the project is:

npx @apify/actors-mcp-server

Configure the client to run that command and provide an Apify API token through APIFY_TOKEN. The package also accepts command-line options; the repository points to npx @apify/actors-mcp-server --help for the available settings.

For remote deployments, configure the client with https://mcp.apify.com. The former /sse endpoint is not supported; clients should use the base URL and Streamable HTTP instead. The README states that the server supports Claude Desktop, Claude.ai, Cursor, Visual Studio Code, and other MCP-compliant clients. A downloadable MCPB bundle and an Apify Tester MCP Client are also provided for trying the service without a manual local setup.

Tools and capabilities

The listed MCP tools include:

• search-actors for finding available Actors.
• fetch-actor-details for inspecting an Actor.
• call-actor for starting an Actor task.
• get-dataset for obtaining Actor dataset output.
• fetch-apify-docs for retrieving Apify documentation.

The server is intended for tasks where an agent must select an existing Apify tool and use its returned data. It does not imply that every Actor exposes the same inputs or output format; those details come from the selected Actor.

Limitations and notes

A valid Apify token is needed for the documented local setup. Hosted access can use OAuth or a bearer token, and the README also describes optional agentic payment methods for running Actors without a conventional Apify API token. Actor availability, cost, input schema, and result quality are governed by the selected Apify Store Actor rather than by the MCP interface itself.

The apify/actors-mcp-server MCP server is not limited to the example scrapers in the README: its purpose is to make Apify Actors available through MCP. However, the client must support the relevant MCP transport and capabilities, and local stdio does not provide all hosted-server features.