# Apify MCP Server

> An MCP server that connects AI agents to 55,000+ Apify Actors for web scraping, data extraction, and automation via a hosted endpoint or CLI.

The Apify MCP Server is an open-source Model Context Protocol (MCP) server that bridges AI agents with Apify's ecosystem of ready-made scrapers, crawlers, and automation tools. It is available as a hosted endpoint at `mcp.apify.com` and as an npm package (`@apify/actors-mcp-server`) for local stdio use. The project is maintained by Apify, a Prague-based company founded in 2015, and is licensed under MIT.

## What It Is

The Apify MCP Server implements the Model Context Protocol to expose Apify Actors as callable tools for AI assistants and agents. An Actor is a containerized, ready-made automation task — examples include scrapers for Google Maps, Instagram, Facebook, and Google Search, as well as a RAG Web Browser for general web retrieval. The server dynamically loads an Actor's input schema and presents it as a structured MCP tool, so the AI agent knows exactly what arguments to pass and what output to expect. The hosted server at `mcp.apify.com` supports OAuth and Streamable HTTP transport, making it connectable from clients like Claude.ai, Cursor, VS Code, ChatGPT, and Grok simply by pasting the URL.

## Tool Architecture and Dynamic Discovery

The server ships with a default tool set and supports dynamic expansion at runtime:

- **`search-actors`** — searches Apify Store for relevant Actors
- **`fetch-actor-details`** — retrieves input schema, README, pricing, and output schema for a specific Actor
- **`call-actor`** — executes any Actor and returns run results
- **`get-actor-run` / `abort-actor-run`** — inspect or stop running Actor jobs
- **`get-dataset-items` / `get-key-value-store-record`** — retrieve structured output from Actor runs
- **`search-apify-docs` / `fetch-apify-docs`** — search and fetch Apify platform documentation

By default the server pre-loads the `actors` category, the `docs` category, and the `apify/rag-web-browser` Actor. Operators can narrow or expand the tool set via URL query parameters (hosted) or CLI flags (stdio), for example `?tools=apify/my-actor` to expose only a single Actor with no discovery overhead.

## Deployment Model

The server supports two deployment paths:

- **Hosted (recommended):** Connect any MCP-compatible client to `https://mcp.apify.com` using OAuth or a Bearer token. Supports the latest features including output schema inference for structured Actor results and rental Actors from Apify Store.
- **Local stdio:** Run `npx @apify/actors-mcp-server` with an `APIFY_TOKEN` environment variable. Suitable for Claude Desktop and command-line integrations. Note that rental Actors and some hosted-only features are not available in stdio mode.

The server also supports a UI mode (`?ui=true`) that enables interactive MCP Apps widget rendering in tool responses.

## Agentic Payments

The server supports three payment paths for agents that operate without a pre-configured Apify account:

- **AGI** (`agi.apify.com`): The recommended path. An agent pays once via x402 or MPP and receives a prepaid, spend-capped Apify API token usable against any Actor.
- **Direct x402**: Pays with USDC on the Base blockchain per request, handled automatically by the `mcpc` CLI client. Limited to Pay Per Event Actors.
- **Skyfire**: Pays with PAY tokens via a Skyfire account. The agent passes a payment token in the Actor's input properties. Also limited to Pay Per Event Actors.

## Update: v0.14.1

The latest release is **v0.14.1**, published on July 31, 2026. The repository (2,420 stars, 206 forks as of the last update) is actively maintained with pushes as recent as August 2026. A notable recent change is the removal of the legacy SSE transport (`/sse` endpoint) in favor of Streamable HTTP — clients previously using `https://mcp.apify.com/sse` must migrate to `https://mcp.apify.com`. The v2 configuration introduced a unified `tools` parameter that supersedes the older `actors` parameter while preserving backward compatibility.

## MCP Client Compatibility

The server is tested against Claude Desktop, Claude.ai (web), Claude Code, ChatGPT, VS Code (Genie), Cursor, OpenCode, Kiro, and the Apify Tester MCP Client. A one-click `.mcpb` bundle file is available for clients that support the MCP bundle format. Docker Hub integration is also maintained via the `docker/mcp-registry` repository.

## Features
- Hosted MCP endpoint at mcp.apify.com with OAuth support
- Access to 55,000+ Apify Actors as MCP tools
- Dynamic Actor discovery via search-actors and fetch-actor-details
- Pre-configured RAG Web Browser Actor for web search
- Streamable HTTP and stdio transport modes
- Configurable tool sets via URL query parameters or CLI flags
- Agentic payments via AGI, x402 (USDC on Base), and Skyfire
- Actor run management: status, logs, abort
- Dataset and key-value store retrieval
- Apify documentation search and fetch
- UI mode for interactive MCP Apps widget rendering
- OAuth and Bearer token authentication
- Docker Hub integration
- Telemetry with opt-out support
- Backward-compatible v2 configuration

## Integrations
Claude Desktop, Claude.ai, Claude Code, ChatGPT, Cursor, VS Code, Grok, Kiro, OpenCode, Apify Store, Apify API, Skyfire, x402 / Base blockchain, AGI (agi.apify.com), Docker Hub, Smithery, MCP Inspector

## Platforms
LINUX, WEB, API, VSC_EXTENSION, CLI

## Pricing
Open Source, Free tier available

## Version
v0.14.1

## Links
- Website: https://mcp.apify.com
- Documentation: https://docs.apify.com/platform/integrations/mcp
- Repository: https://github.com/apify/actors-mcp-server
- EveryDev.ai: https://www.everydev.ai/tools/apify-mcp-server