Apify MCP Server: Connect AI Agents to Thousands of Web Scrapers via MCP
The Apify MCP server enables your AI agents to extract data from social media, search engines, maps, e-commerce sites, or any other website using thousands of ready-made scrapers, crawlers, and automation tools available on the Apify Store.
At a glance
- What is it?
- The Apify MCP Server exposes thousands of ready-made web scrapers, crawlers, and automation tools from Apify Store as callable tools to any MCP-compatible AI agent. It runs as a hosted endpoint at mcp.apify.com supporting OAuth, or locally via stdio with an API token, and supports agentic payments for per-Actor billing.
- Who is it for?
- The Apify MCP Server is the right tool when an AI agent needs to extract real data from websites, social media, or search engines without the engineer building a scraper from scratch. It is not the right choice for agents that run entirely locally or offline, since Actors execute in Apify's cloud and require an Apify account and API token.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 1 day ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 29, 2026, and from our analysis. They are not legal advice.
Editorial analysis
What the Apify MCP Server Does and Who It Is For
The Apify MCP Server bridges the Model Context Protocol and Apify's cloud infrastructure. An MCP-compatible AI agent, such as Claude Desktop or a Claude Code agent, can call any Actor in the Apify Store as if it were a local tool. The agent does not manage scraping infrastructure; it calls a named Actor with input parameters and receives structured output.
The README gives concrete examples of what this enables: extracting posts from Facebook pages using the apify/facebook-posts-scraper Actor, extracting business contact details from Google Maps with lukaskrivka/google-maps-with-contact-details, scraping Google search results with apify/google-search-scraper, collecting Instagram content with apify/instagram-scraper, browsing the web and returning Markdown content with apify/rag-web-browser, and fetching any URL as plain text or HTML with apify/web-fetch.
The intended users are AI engineers and agent developers who want web data capabilities in their agents without writing or maintaining scrapers. The package name is @apify/actors-mcp-server on npm. Current version is 0.16.1.
Two Connection Modes: Hosted HTTPS and Local stdio
The server operates in two modes from the same codebase.
The hosted mode at https://mcp.apify.com is the recommended path for most users. It supports OAuth and the Streamable HTTP transport. Clients that support OAuth (including Claude.ai and Visual Studio Code) can connect using just the URL, with no manual token configuration. Clients that do not support OAuth include the Authorization: Bearer <APIFY_TOKEN> header in their requests.
The stdio mode runs the server locally as a command-line process. Set the MCP client server command to npx @apify/actors-mcp-server and set the APIFY_TOKEN environment variable to your Apify API token:
npx @apify/actors-mcp-serverThe .env.example file in the repository shows the required environment variable:
APIFY_TOKEN=The stdio mode does not support output schema inference for structured Actor results. That feature is only available in the hosted mode. Engineers who need structured JSON output from Actors should use the hosted endpoint.
The legacy SSE endpoint at https://mcp.apify.com/sse has been removed. Existing clients must drop the /sse suffix and connect to https://mcp.apify.com.
Supported MCP Clients and Tested Integrations
The server is compatible with any MCP client that follows the Model Context Protocol specification. The README lists tested clients: Claude Desktop, Claude.ai (web), ChatGPT, VS Code (Genie), Cursor, OpenCode, Kiro, and the Apify Tester MCP Client.
The Apify Tester MCP Client is a chat-like interface built specifically for testing Apify MCP servers, available as an Actor in the Apify Store at apify.com/jiri.spilka/tester-mcp-client. It allows developers to experiment with the server without configuring a separate MCP client.
A one-click install option is available as an MCPB file (formerly the Anthropic Desktop extension format). The MCPB file is available in the GitHub releases at the repository's latest release URL.
Tested versions follow the MCP SDK version 1.30.0 as listed in package.json. Node.js 22.0.0 or higher is required. The package manager is pnpm 11.11.0.
Agentic Payments: AGI Tokens and x402
The server supports three agentic payment mechanisms for running Actors that require credits.
The recommended method is AGI tokens: buy a token from AGI and the server automatically applies it to run any Actor. This is designed for agents that autonomously decide which Actor to call without human involvement in the billing step.
Direct x402 payment is available for Pay Per Event Actors. The agent pays per-request using the x402 payment protocol.
Skyfire is a third payment option listed in the README. The README does not describe Skyfire's mechanism in detail beyond naming it as a supported payment option.
For agents that run within an organization's Apify account with a normal API token, Actor usage is billed against that account's standard Apify credit balance. The agentic payment options are for scenarios where the agent needs to pay autonomously without a pre-loaded account.
Architecture: How the Server Translates Actors to MCP Tools
Each Apify Actor has an input schema that describes its parameters. The MCP server reads those schemas and exposes each Actor as an MCP tool with the Actor's input schema as the tool's parameter definition. When an AI agent calls the tool, the server submits the Actor run to Apify's infrastructure and waits for the result.
The hosted server adds output schema inference: it examines the Actor's output dataset and infers a structured schema for the result, so the agent receives typed JSON rather than raw data. This inference step is not available in the stdio mode, which returns the Actor output as-is.
The Dockerfile shows a two-stage build: a builder stage using Node 24 Alpine compiles the TypeScript source with pnpm, and a runtime stage deploys only the production dependencies using pnpm deploy --legacy --filter. The server entry point is dist/stdio.js for the stdio mode.
Limitations and Comparison with Playwright MCP
The Apify MCP Server requires an Apify account and API token. Every Actor run consumes Apify credits. Developers who need web scraping without a cloud account or billing relationship will need a different approach.
Playwright MCP, maintained by Microsoft, is an alternative MCP server for web automation. The fundamental difference: Playwright MCP runs a local headless browser instance, so it can access any URL but requires Chromium installed locally and handles only what a browser can access manually. The Apify MCP server routes requests to cloud-based Actors that are pre-built for specific sites and include anti-bot handling, proxy rotation, and login flows for platforms like Facebook and Instagram that actively block generic browser automation.
The Apify MCP Server also does not guarantee Actor availability. Actors in the Apify Store are maintained by their individual authors; an Actor can be updated, deprecated, or removed by its maintainer independently of the MCP server.
The repository is licensed under MIT. The last push was on 2026-09-27.
Editorial conclusion
The Apify MCP Server is the right tool when an AI agent needs to extract real data from websites, social media, or search engines without the engineer building a scraper from scratch. It is not the right choice for agents that run entirely locally or offline, since Actors execute in Apify's cloud and require an Apify account and API token. Clients that previously connected to https://mcp.apify.com/sse must migrate to https://mcp.apify.com; the SSE endpoint has been removed.
Frequently asked questions
Is Apify an MCP?
Apify is not itself an MCP implementation; it is a platform for web scraping and automation. The apify/apify-mcp-server repository provides an MCP server that connects MCP-compatible AI agents to Apify's platform, allowing agents to call Apify Actors as MCP tools.
How do I install the Apify MCP server?
Run npx @apify/actors-mcp-server with the APIFY_TOKEN environment variable set to your Apify API token for the local stdio mode. For the hosted mode, connect your MCP client to https://mcp.apify.com using OAuth or by including an Authorization: Bearer header. Detailed setup instructions are at mcp.apify.com.
What is the Apify MCP server?
The Apify MCP server is an open-source TypeScript server that implements the Model Context Protocol and exposes Apify Actors as callable tools for AI agents. It runs as a hosted endpoint at mcp.apify.com or locally via stdio, and supports Claude Desktop, Claude.ai, Cursor, VS Code, and other MCP clients.
Is the Apify MCP server free?
The server software is MIT-licensed and free. Running Actors through the server consumes Apify credits, which are billed to your Apify account. Some Actors are free to run on Apify's free tier; others require a paid plan. The agentic payment options (AGI tokens, x402) allow per-run billing without a pre-loaded account balance.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/apify-apify-mcp-server)