The Apify MCP Server provides a TypeScript-based platform to run Apify Actors as MCP tools, exposing authentication, multiple payment options, and dynamic tool discovery. It ships with a hosted MCP server and stdio option, plus examples and a rich tooling set.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
The Apify MCP Server enables AI agents to use Apify Actors as tools through the Model Context Protocol (MCP). It supports hosted usage at mcp.apify.com and stdio-based local runs, and integrates with OAuth to connect clients like Claude or VS Code. The project provides dynamic tool discovery, an initial set of built-in tools, and guidance for interacting with Actors, runs, and storage.
How it works
The server loads Actors from Apify Store as MCP tools by reading each Actor's input schema and creating a corresponding MCP tool. It supports dynamic tool discovery, and clients interact via MCP-compliant calls. Payments for Actor runs can be managed via AGI, Direct x402, or Skyfire, with setup steps described for each path. The server can be accessed via HTTPS endpoint or stdio, and includes a configuration for multiple MCP servers when using Skyfire.
Getting started
Two main modes:
- HTTPS Endpoint (mcp.apify.com): connect via OAuth or Authorization: Bearer <APIFY_TOKEN> header.
- Example endpoint: https://mcp.apify.com
- Standard Input/Output (stdio): run locally with npx @apify/actors-mcp-server and set APIFY_TOKEN in the environment.
- Command example in README: See detailed steps under Quickstart for how to run and connect.
See the README for full instructions and references to the Apify documentation for MCP setup.
Recent releases
- v0.14.1 (2026-07-31): Features include Add outputSchema to report-problem and get-actor-log.
- v0.14.0 (2026-07-29): Features include Serve 2026-07-28 MCP requests from a stateless adapter.
- v0.13.0 (2026-07-27): Features include breaking change; delete add-actor implementation (stateless migration).
- v0.12.0 (2026-07-23): Features include Attribute Apify AI runs with the APIFY_AI origin.
- v0.11.7 (2026-07-21): Features include Storage as resources.
Traction
2517 stars on the repository.
Behind the repo
The README describes hosted MCP server at mcp.apify.com and mentions compatibility with Claude Desktop, Claude.ai, Cursor, VS Code, and other clients. The project is MIT licensed and implemented in TypeScript.
Caveats
License: MIT. Creation date: 2025-01-02. Last push: 2026-08-04. Open issues: 105.