Setup Guide

Connect Claude to
52,000 scrapers

Claude can't visit websites and pull structured data on its own. Apify can. Connect the two once, and you can ask for a lead list, a competitor's posts, or a pile of reviews in plain English, and get a real dataset back.

4 Steps
~10 min Setup time
52,000+ Tools available
$5/mo Free credit

What is Apify?

Apify is a marketplace of ready-made scrapers. Someone has already built and maintained the tool that pulls data out of Google Maps, Instagram, TikTok, LinkedIn, Amazon, or any regular website. You just point it at a target and it hands back clean, structured data.

Each of those tools is called an Actor. There are over 52,000 of them in the Apify Store. You don't install any of them. They run on Apify's servers, not your laptop.

So what does connecting it to Claude do? It turns those 52,000 tools into things Claude can pick up and use by itself. You ask in English, Claude works out which Actor to run, runs it, gets the data back, and then does the thinking part: sorting it, summarising it, spotting the pattern. Scraping and analysis in one move, instead of a CSV you have to go and deal with.

What you'll have after this
An Apify account
$5 of credit every month, no card required
Claude & Cowork connected
One connector, works in the app and on the web
Claude Code connected
One command in the terminal
4 prompts that work
Leads, competitor posts, reviews, teardowns
A way to find any Actor
Ask Claude to search the store for you
A clear cost model
What's free, what isn't, and how to check first
Do this in the next 10 minutes

Here's where you're going

This is the prompt you'll paste once you're connected. It's the whole point of the four steps below. Read it now so the setup has a destination.

Prompt
Use the Google Maps Scraper on Apify to find [plumbers] in [my suburb].
Cap it at 25 results.

Give me a table: business name, phone, website, star rating, review count.
Then tell me which three have the weakest online presence, and why.

Ten minutes from now that returns a real, sorted list of businesses in your market, with the analysis already done.

Before you start: You need a Claude account. The free plan works, though it's limited to one custom connector. Apify's free plan gives you $5 of credit a month and doesn't ask for a card. Steps 2 and 3 are alternatives, not both: pick Claude chat/Cowork, or Claude Code, or do both if you use both.
1

Create a free Apify account

This is the account that actually runs the scrapers. Sign up with Google, GitHub, or an email address. There's no credit card step.

console.apify.com/sign-up

That's the whole step. Your free plan includes $5 of platform credit every month, which is enough to pull hundreds of Google Maps listings while you work out whether this is useful to you.

Worth knowing: the $5 resets monthly and doesn't roll over. If you run out, the free plan pauses until the next cycle rather than charging you.
2

Connect it to Claude and Cowork

This is the one to do if you use Claude in the desktop app, in the browser, or in Cowork. It's the same connector for all three. Set it up once and it follows you across them.

In Claude, go to Settings → Connectors and click Browse connectors. Search for Apify and add it.

Claude opens a browser tab asking you to sign in to Apify and authorise the connection. Click allow. That's it. No API key to copy or paste anywhere.

To check it worked, start a new chat and ask Claude: "What Apify tools do you have?" It should list things like searching the store and running Actors.

On the free Claude plan? You get one custom connector. If you've already used it on something else, you'll need to remove that one first, or this is a good argument for the Claude Code route in Step 3, which doesn't count against it.
3

Connect it to Claude Code

Skip this if you don't use Claude Code. Step 2 is enough on its own. If you do, it's one command in your terminal.

Terminal
claude mcp add --transport http apify https://mcp.apify.com --scope user

--scope user makes Apify available in every project on your machine. Leave it off and it only works in the folder you're currently in.

Now start Claude Code and run this to sign in:

In Claude Code
/mcp

Pick apify from the list, choose Authenticate, and approve it in the browser tab that opens. The same /mcp command shows you a green connected status afterwards.

4

Pull your first real dataset

Don't test it with something abstract. Test it with your own market, so the output is worth keeping. Paste this and swap the trade and suburb for yours:

Prompt
Use the Google Maps Scraper on Apify to find plumbers in Parramatta, NSW.

Cap it at 25 results so we don't burn credit.

For each one, give me: business name, phone, website, star rating, and
review count. Skip any without a phone number.

Show it as a table, sorted by review count, highest first. Then tell me
which three look like the weakest online presence and why.

Claude will find the right Actor, ask you to approve the run, then hand back a table. The last line is the bit that matters. That's the part a bought lead list can't do.

Always cap your first run. "Cap it at 25 results" is the single most useful habit here. Actors will happily pull 5,000 rows and eat your monthly credit in one go. Start small, check the data's any good, then scale up.

What it actually costs

This is the part most guides skip, so here it is straight. There are two layers, and only the first one is covered by your free plan.

1
Your Apify plan
Free = $5 credit/month, no card. Paid plans start at $29/month. This covers the compute the scraper uses.
2
The Actor's own price
Set by whoever built it. Some are free. Some charge per result or per event, on top of layer 1.

Every Actor's price is shown on its own page in the store before you run anything, so you're never guessing. Connecting Apify to Claude costs nothing extra by itself. There's no fee for the connection, only for the runs.

Easiest way to stay honest with yourself, just ask before you commit:

Prompt
Before you run anything: what does this Actor charge, and roughly what
will 25 results cost me? If it's pay-per-result, say so and wait for me
to confirm.

What people actually use it for

The store has 52,000 Actors, which is useless as a starting point. These are the jobs that come up over and over for a business owner:

google-maps Local lead lists
reviews What customers complain about
instagram Competitor content teardowns
tiktok What's working in your niche
linkedin Prospect and hiring research
amazon Product and price tracking
website-crawler Read a whole site at once
google-search Rankings and SERP data
facebook-ads Competitor ad libraries
youtube Video and channel data
The honest version: this is worth setting up if you regularly need data about a market you can't get from inside your own tools (competitors, prospects, reviews, pricing). If you don't, a $5/month scraping credit won't change your business, and that's a fine conclusion to reach.

Prompts worth stealing

Every one of these does the scrape and the thinking. Swap the bracketed bits for yours and paste them in.

1. Competitor content teardown

Find out what's working for someone else before you post again.

Prompt
Use the Instagram Scraper on Apify to pull the last 30 posts from
[@competitorhandle]. Get caption, likes, comments, and post URL.

Then: rank them by engagement, show me the top 5, and tell me what those
5 have in common that the bottom 5 don't — hook style, format, topic,
length. Be specific, not generic.

Finish with 3 content angles they haven't covered that I could own.

2. Mine your competitors' bad reviews

Their 1-and-2-star reviews are your positioning, written by their customers.

Prompt
Use Apify to pull Google reviews for the top 5 [dentists] in [Bondi].
Up to 100 reviews each, and tell me the cost before you run it.

Filter to 1, 2 and 3 star reviews only. Find the 5 complaints that come
up most often across all of them, with a real quote for each.

Then write 3 positioning angles a new [dentist] could use that speak
directly to those complaints. Plain language, no marketing waffle.

3. Read a competitor's entire website

Their offer, pricing and promises, side by side with yours.

Prompt
Use the Website Content Crawler on Apify to crawl [competitor.com].
Max 25 pages.

Summarise: what they sell, their pricing if it's public, their guarantees,
and exactly who they say they're for.

Then crawl [mysite.com] the same way and give me a side-by-side table of
where their offer is stronger than mine. Be blunt — I want the gaps, not
reassurance.

4. A lead list that's already qualified

The filtering is the value. Anyone can pull 500 names.

Prompt
Use the Google Maps Scraper on Apify to find [cafes] in [Newtown, NSW].
Cap at 50 results.

Then filter to only the ones that look like they need help: no website,
or fewer than 20 reviews, or rated under 4.0.

Give me a table with name, phone, website, rating, review count, and a
one-line note on why they made the list. Save it as a CSV I can open
in Excel.

How to find the right Actor

You don't browse 52,000 tools. You ask Claude to browse them, because searching the store is one of the things the connection gives it.

Prompt
Search the Apify store for Actors that can [scrape LinkedIn company pages].

Show me the top 3 by popularity and rating. For each one: what it does,
what it charges, and what it needs as input.

Then recommend one and tell me why — including why you didn't pick the
other two.

You can also browse it yourself at apify.com/store and sort by users. The ones with hundreds of thousands of runs are popular for a reason.

What separates a good prompt from a bad one

Every Actor needs the same four things. Miss one and you get a mess back, so say all four up front:

Then add the fifth thing: tell it what to think about the data. "Sort it" is a scrape. "Tell me which three are weakest and why" is the reason you connected an AI to a scraper instead of downloading a CSV.
What this quietly replaces: the $200 lead list you bought once and never called through. The VA copy-pasting business names into a spreadsheet. The competitor tracker you set up in a Google Sheet and abandoned in week three. The review-monitoring tool with a monthly fee you forgot you were paying. None of that needed a person or a subscription. It needed the data pulled once, thought about properly, and handed back as an answer.