How to Connect Apify to Claude Desktop via MCP Server (2026)

Contains affiliate links. Disclosure.

Connect Apify to Claude Desktop via the official MCP server. Claude can run Actors from chat, browse the Apify Store, and read datasets, no code required. You ask; Claude scrapes. If you do not have the app yet, you can try Claude free for a week, then get an Apify account and follow the steps below.

What MCP is and why it matters

Model Context Protocol (MCP) is an open standard (originated by Anthropic) that lets AI apps call external tools. Claude Desktop supports MCP natively, and so do Claude Code, ChatGPT, VS Code, and Cursor, so the same Apify connector works across them. When you add the Apify MCP server, Claude gains tools to:

• Run Actors (e.g., Google Search Scraper, Website Content Crawler)
• Search the Apify Store
• Fetch dataset results from past runs

Claude decides when it needs live data, calls the tool, waits for the result, and uses it in its reply. No copy-paste, no manual API calls. For hosted vs local setup, alternatives, and a deeper look at the architecture, see the Apify MCP server guide for Claude Desktop.

Apify's official MCP server

Apify offers two ways to connect. The hosted server at https://mcp.apify.com/?fpr=use-apify is the recommended path in 2026: you add it as a remote MCP connector, authenticate with OAuth on first use (no token in a file), and it stays auto-updating. This guide covers the local option, which is handy when you want the token to stay in your own environment or your client cannot add a remote connector.

@apify/actors-mcp-server is Apify's official npm package for the local route. It runs as a subprocess. Claude spawns it, sends JSON-RPC over STDIN, receives tool results over STDOUT (the stdio transport). You configure which Actors are allowed via the --actors flag.

Prerequisites

• Node.js (18+) for npx
• Claude Desktop, free or paid. The app is available to free and paid users; heavier MCP use and connectors are smoother on a paid plan (Pro or Max).
• Apify account: sign up
• APIFY_TOKEN from Apify Console, under Settings, Integrations, API tokens

Step-by-step setup

Step 1: Install Node.js

If you don't have Node.js, install from nodejs.org. Verify:

node --version   # v18+
npx --version    # 10+

Step 2: Get your Apify API token

1. Log into Apify Console.
2. Click your profile → Settings → Integrations.
3. Under API tokens, create a new token or copy an existing one.
4. Store it somewhere safe. You'll put it in claude_desktop_config.json.

Step 3: Edit claude_desktop_config.json

Config locations:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Create or edit the file. Add the Apify MCP server:

{
  "mcpServers": {
    "apify": {
      "command": "npx",
      "args": [
        "-y",
        "@apify/actors-mcp-server",
        "--actors",
        "apify/google-search-scraper,apify/website-content-crawler"
      ],
      "env": {
        "APIFY_TOKEN": "apify_api_YOUR_TOKEN_HERE"
      }
    }
  }
}

Replace apify_api_YOUR_TOKEN_HERE with your real token. The --actors list limits which Actors Claude can run. Start with 1–2 to avoid context bloat.

Step 4: Restart Claude Desktop

Quit Claude completely (not just close the window). Reopen it. You should see a plug icon in the UI if MCP connected. If not, check the config path and JSON syntax.

Step 5: Test from Claude

Ask Claude: "Scrape the top 10 Google search results for 'web scraping tools 2026' and summarize them."

Claude will call the Google Search Scraper tool, wait for results, and return a summary. If it works, you're done.

What you can do once connected

Capability	Example prompt
Run Actors	"Scrape the first 5 results from Google for 'best laptops'"
Crawl a site	"Extract the main content from https://example.com/blog/post"
Browse Store	"What SEO Actors are available on Apify?"
Read datasets	"Summarize the dataset from my last Google Search run"

Be specific. "Scrape 10 results" is better than "scrape everything." Large runs cost CUs and can hit Claude's context limit. You can restrict Actor scope further by passing --actors with only the Actor IDs you want exposed. Fewer tools mean less prompt bloat and faster tool resolution.

Example: SERP scrape via Claude

You: "Use the Google Search Scraper to get the top 10 organic results for 'Apify alternatives' and list the URLs and titles."

Claude: Calls the tool with queries: ["Apify alternatives"], maxPages: 1, resultsPerPage: 10. When the dataset returns, Claude formats it and lists URLs/titles.

Output: Structured list you can copy or refine. No manual Actor runs.

Real-World Use Cases

Once connected, the most powerful patterns combine Claude's reasoning with Apify's library of 30,000+ pre-built Actors.

Market research and sentiment analysis

Instead of manually searching social media, ask Claude to analyze brand sentiment across networks.

Prompt: "Run the clockworks/tiktok-comments-scraper for videos mentioning 'brand name'. Summarize the top 500 comments, identifying the main complaints and praises."

Claude triggers the TikTok scraper, reads the JSON dataset, and generates a qualitative report that would take hours to compile manually.

Lead generation and outreach

Prompt: "Use the Google Maps Scraper to find dental clinics in Austin, TX. Then crawl the top 10 clinic websites and extract the head dentist's name and contact info."

Claude chains two Actors autonomously (Maps Scraper output feeds into the Website Crawler), then returns a structured contact list.

Competitor price monitoring

Prompt: "Run the Amazon Product Scraper on these 5 competitor ASINs. Create a Markdown table comparing current prices, ratings, and stock status. Highlight any price drops since last week."

Claude fetches live Amazon data and formats it as a readable table, instant pricing intelligence without leaving chat.

Content research and SERP analysis

Prompt: "Search Google for 'best vllm alternatives 2026' and give me the top 10 results with their URLs, titles, and word counts."

Claude runs the Google Search Scraper, parses the dataset, and returns a structured research brief in seconds.

Apify MCP vs Bright Data MCP vs Firecrawl MCP

Attribute	Apify MCP	Bright Data MCP	Firecrawl MCP	Best For
Data source	30,000+ Actors (SERP, maps, stores)	SERP, scraping APIs	URLs to Markdown, search	Apify: breadth
Pricing	Usage-based CUs	Per-request / plan	Per-page / plan	Depends on volume
Setup	APIFY_TOKEN, --actors	API key	FIRECRAWL_API_KEY	All similar
Claude integration	Run Actors, Store, datasets	Scrape, SERP tools	Scrape, map, extract	Apify: full platform
Cold start	Varies by Actor	Low	Low	Firecrawl: fastest

Winner for general scraping: Apify, one MCP with many data sources. Best for simple URL-to-Markdown: Firecrawl. See Firecrawl MCP setup for comparison.

Troubleshooting

Issue	Fix
Token not found	Put APIFY_TOKEN in the env block. Restart Claude. Check for typos.
Server not starting	Run npx -y @apify/actors-mcp-server in a terminal. Fix any Node/npm errors.
Tools not visible	Restart Claude. Check Cursor/Claude MCP settings if using an IDE.
Timeout errors	Reduce Actor input (fewer URLs, lower maxItems). Some Actors need minutes.
Payload too large	Even with large context windows (current Claude models reach 1M tokens), huge datasets waste tokens. Instruct Claude to use limit: 10 or similar.

See Make.com Claude MCP for another MCP pattern if you need workflow triggers.

Frequently asked questions

Below are the exact questions people ask when wiring Apify MCP into Claude Desktop, including how to connect Apify to Claude and where to put your Apify token.

Start with one Actor

Add apify/google-search-scraper first. Test with a simple query. Add more Actors only when you need them.

Get your APIFY_TOKEN

Frequently Asked Questions

Install Node.js 18+, copy your APIFY_TOKEN from Apify Console → Settings → Integrations, add the @apify/actors-mcp-server block to claude_desktop_config.json (see Step 3 in this guide), list allowed Actors with --actors, then fully restart Claude Desktop.

Apify MCP is the Model Context Protocol server package @apify/actors-mcp-server. It exposes Apify tools (run Actor, search Store, read datasets) to Claude Desktop over JSON-RPC so Claude can scrape and retrieve data without you pasting API responses.

Put APIFY_TOKEN inside the env object of your apify MCP server entry in claude_desktop_config.json, the same file path as Step 3. Never commit this file to a public repo.

Only Actor IDs you pass to --actors (e.g. apify/google-search-scraper,apify/website-content-crawler). Claude cannot invoke Actors outside that allowlist. Add more IDs only when you need them.

Actors can take minutes for large jobs. Reduce input (fewer URLs, lower maxItems). Check Run details in Apify Console for actual duration. Some Actors have cold starts.

Apify MCP gives access to 30,000+ Actors (SERP, maps, e-commerce, social). Firecrawl MCP focuses on URL scraping, search, and extraction. Use Apify for breadth; Firecrawl for simple scrape-to-markdown.

Claude Desktop is available to free and paid users. You can experiment with MCP without a paid plan, but connectors and heavier use are smoother on a paid plan (Pro or Max). Check Anthropic's current plan features.