Apify MCP Server lets an AI assistant use any Actor from Apify Store as a tool. Actors are ready-made scrapers, crawlers, and automation programs, for example ones that extract Facebook posts, pull contact details from Google Maps, scrape Google Search results, or collect Instagram profiles and comments. The server loads an Actor's input schema and turns it into an MCP tool, so the model knows what arguments to pass and what comes back.
Because Apify Store is large, the server supports dynamic tool discovery: the agent searches for Actors, reads their details, and adds them as tools during a session instead of relying on a fixed list. By default it loads Actor search, details, and calling, Apify documentation search, and two Actors (a RAG web browser that searches and scrapes the top results, and a web fetcher that returns a page as Markdown, text, HTML, or links). Further tool categories cover Actor runs, datasets and key-value stores, saved tasks, schedules, builds, Actor source files, and read access to the Apify API, and each can be selected by category or by name.
The server runs hosted, connecting over Streamable HTTP with OAuth or a bearer token, or locally over stdio. The hosted version adds output schema inference for structured Actor results. Tested clients include Claude Desktop, Claude.ai, Claude Code, ChatGPT, VS Code, Cursor, OpenCode, and Kiro. An agent can also pay for Actor runs without an Apify account: AGI mints a prepaid, spend-capped token for an x402 or MPP payment, and direct x402 (USDC on Base) and Skyfire PAY tokens settle per run for Pay Per Event Actors.
Features
- Actors as tools: any Apify Store Actor becomes an MCP tool generated from its input schema
- Dynamic tool discovery: the agent searches Apify Store and adds new Actors as tools mid-session
- Default tool set: Actor search, details, and calling, Apify docs search and fetch, a RAG web browser, and a web fetcher
- Runs and storage: inspect, list, and abort Actor runs, read their logs, and fetch dataset items and key-value store records
- Tasks and schedules: create, update, and publish saved Actor tasks, and manage cron schedules that run Actors and tasks
- Builds and source: build Actor versions, check build status and logs, and read an Actor version's files
- Apify API access: search the API reference, inspect operations, and send GET requests
- Tool selection: load tools by category or by name, down to a single Actor with no discovery tools
- Agentic payments: pay for runs without an account through AGI tokens, direct x402 in USDC on Base, or Skyfire PAY tokens
- Transports: a hosted Streamable HTTP endpoint with OAuth, or a local stdio server
- MCP Apps UI: an optional mode returns interactive widget responses from tools such as Actor search
- Tool annotations: read-only and open-world hints on every tool to describe its behavior to clients
