Apify MCP Server – A bridge that lets AI assistants run Apify Actors

What it is

The Apify Model Context Protocol (MCP) server (mcp.apify.com) is a hosted service that implements the Model Context Protocol – a standard way for large‑language‑model (LLM) agents to call external tools. In this case the tools are Apify Actors, which are ready‑made web‑scraping, data‑extraction and automation scripts that run on the Apify platform.

Who would use it

• LLM‑powered assistants (Claude, Claude Desktop, Cursor, VS Code extensions, ChatGPT, etc.) that need to fetch live web data or run a custom automation step.
• Developers building autonomous agents who want a plug‑and‑play catalogue of 8 000+ Actors without writing HTTP code themselves.
• Teams that want to bill agents per‑run via the built‑in agentic payments (AGI token, x402 USDC, or Skyfire PAY tokens).

Core capabilities

Feature	What it does	Why it matters
Dynamic tool discovery	The server reads an Actor’s input schema and automatically creates an MCP tool (e.g., call-actor, search-actors). The LLM knows exactly which arguments to pass.	No hard‑coding of tool signatures; new Actors become instantly usable.
HTTPS endpoint with OAuth	Clients call https://mcp.apify.com with an Authorization: Bearer <APIFY_TOKEN> header or via OAuth flow.	Secure, token‑based access that works from cloud‑hosted agents and desktop clients.
StdIO mode	Run the server locally with npx @apify/actors-mcp-server and pipe JSON messages over stdin/stdout.	Useful for offline or CLI‑based integrations (e.g., Claude for Desktop).
Agentic payments	Pay for Actor runs without an API token using:

• AGI – prepaid Apify token minted after an x402 or MPP payment (recommended).
• Direct x402 – USDC on Base blockchain, handled by the mcpc client.
• Skyfire – PAY‑token system. | Enables “pay‑as‑you‑go” AI agents that can run paid Actors even when the user has no Apify account. | | Rich tool set | Beyond call-actor, the server offers helpers for searching the Apify Store, fetching docs, managing datasets, key‑value stores, schedules, builds, etc. | Gives agents full CRUD access to the Apify platform, not just scraping. | | Telemetry & reporting | Optional telemetry tools (report-problem, dev) let agents send usage data back to Apify. | Helps improve the service and debug client issues. |

How to get started

1. Pick a client – Claude Desktop, VS Code Genie, the open‑source mcpc CLI, or the web‑based Apify Tester MCP Client.
2. Connect – Use the URL https://mcp.apify.com and provide either an OAuth flow or an APIFY_TOKEN (or an AGI‑generated token).
3. Discover tools – Call search-actors or fetch-actor-details to list available Actors and retrieve their schemas.
4. Run a tool – Example: call-actor actor:="apify/rag-web-browser" input:{"query":"latest AI news","maxResults":3}.
5. Pay (optional) – If you don’t have an API token, obtain an AGI token or configure x402/Skyfire as described in the README.

Example use‑case

An autonomous research assistant receives the prompt “Summarise the top‑3 recent articles about quantum‑computing breakthroughs.” It:

1. Calls search-actors → discovers apify/rag-web-browser.
2. Calls call-actor with the query and maxResults:3.
3. Receives the scraped article texts, summarises them, and returns the answer to the user. All of this happens through a single MCP request/response cycle.

---

Bottom line – The Apify MCP Server is a real, production‑grade bridge that lets LLM agents treat Apify’s massive library of web‑automation Actors as native tools, with built‑in authentication, payment options, and a rich management API.