MIT-licensed MCP server, hosted or local, that lets agents search and run Apify scraping Actors, with 29 documented tools and a 30 req/sec cap.
Agents that need web data — scraping, crawling and page fetching — rather than SaaS write actions. The server turns the Apify Store into callable tools, with search-actors and fetch-actor-details for discovery and call-actor for execution, so an agent can find a scraper at runtime instead of you wiring one per site. Good fit for research assistants and data pipelines, and because the server is MIT-licensed you can read exactly what it exposes and run it locally over stdio for development.
Teams wanting broad SaaS tool coverage should skip it: this is one vendor's platform, so there is no Slack, Salesforce or Gmail action here and the tool surface is Apify Actors plus Apify API management. The defaults are deliberately thin — only the actors and docs categories plus apify/rag-web-browser, apify/web-fetch and report-problem load by default, and the README warns the default set may change, so production use requires pinning the tools parameter yourself. Anyone planning to self-host the stdio server for everything will hit gaps: the docs state rental Actors work only with the hosted server, and newer features such as output schema inference are not available locally. Running Actors costs Apify platform usage regardless of the free MCP code.
Pros
Cons
The server code is MIT-licensed and public at github.com/apify/actors-mcp-server, and you can run it locally with npx -y @apify/actors-mcp-server. Four tools — search-actors, fetch-actor-details, search-apify-docs and fetch-apify-docs — work without an account. Actually running Actors draws on Apify platform usage, and the Free plan includes $5 of usage a month, blocking access until the next cycle once spent.
Yes. For the hosted server at mcp.apify.com the docs call OAuth the recommended path: on first connection a browser opens so you sign in to Apify and approve access. The alternative is an Authorization: Bearer APIFY_TOKEN header, which is what code-based clients use. Local stdio runs read the APIFY_TOKEN environment variable instead. Running Actors or reading storage and run data always requires authentication.
Apify documents 30 requests per second per user on the hosted server, and exceeding it returns a 429 response. Separately, the call-actor tool accepts a waitSecs value between 0 and 45, defaulting to 30 seconds, which caps how long a call blocks waiting for an Actor run. Beyond that, Apify advises monitoring API usage in Apify Console to stay within your plan's usage limits.
The docs tool table lists 29 tools spanning actors, docs, runs, storage, tasks, schedules and dev categories, plus the Actors you choose to load as tools. By default only the actors and docs categories load, along with apify/rag-web-browser, apify/web-fetch and report-problem. The README warns the default may change over time, so set the tools parameter explicitly for anything running in production.
Facts sourced from the vendor’s own site, last verified 2026-10-10.
The Weekly Source
One email a week. New listings, repriced tools, and the ones we would skip - including this one if its pricing changes.
