github-apify-mcp-server

---
name: github-apify-mcp-server
description: "Configure or use Apify MCP Actor discovery and bounded website-data extraction with verified input schemas, run costs, result validation, and current transport settings."
metadata:
  repository: "apify/apify-mcp-server"
---

# Apify MCP

Use Apify MCP to select and run a specific Actor for an authorized data task. A working connector does not make every Actor free, appropriate, or authorized for a large crawl.

1. Define source domains, allowed pages, desired fields, date range, maximum items, and intended use. Prefer existing datasets or a small sample when those satisfy the request. Identify whether the source requires an account or contains personal information before choosing a crawler.
2. Read [upstream evidence](references/upstream.md), refreshing current tools, Actor documentation, prices, and transport before setup, upgrades, or commercial use. The README recommends the hosted Streamable HTTP endpoint at mcp.apify.com; the legacy /sse endpoint has been removed.
3. Discover the relevant Actor and inspect its input schema, pricing model, output fields, and limits. Translate the user's scope into explicit URL/item/page and run-budget settings. OAuth, API tokens, x402, and other payment options are different authorization paths; do not buy credits or authorize payments from a research request.
4. Execute only the bounded authorized run, retaining its run ID, exact input, timestamp, and dataset reference. Keep credentials out of logs and output. If the Actor fails or reaches a cap, inspect its status before retrying; avoid launching duplicate billable runs or broadening collection silently.
5. Validate extracted rows against a source sample. Check missing fields, duplicate pages, dates, currencies, pagination coverage, and whether content came from the intended domain. Report partial results and costs instead of treating a completed run as proof of complete coverage.

Deliver a source-linked table or export with field definitions, provenance, run limits, and observed completeness. Respect source terms and access controls; research does not authorize bulk scraping or outreach to collected contacts.

Examples:
- “Extract product names and prices from these ten public pages.” Bound the run to that list and validate samples.
- “My old Apify MCP connection fails.” Inspect transport configuration and migrate the obsolete /sse URL if that fix is authorized.