Scrape Phones From Website

Give it a company website, get back every phone number published on it.

What it does

• Crawls a company website and extracts every published phone number — page text, tel: links, and embedded markup.
• Covers the homepage plus contact, about, team, support and location pages.
• Returns switchboard, sales and office numbers rather than personal mobiles. You reach a front desk, not a decision-maker’s handset.
• Takes a website URL or a bare domain.

If you need the number of a specific person, this is the wrong tool. find_work_phone returns their office line and find_mobile_number returns their personal mobile — both 12 credits, both from a LinkedIn URL.

Use cases

• Build a call list for local businesses. For trades, clinics and restaurants the switchboard is the intended way in.
• Cover accounts with no profile-based match. Where a LinkedIn lookup fails, the published number usually still exists.
• Add a cheap phone step before an expensive one. At 0.5 credits it is worth trying before spending 12 on find_work_phone.
• Verify a business is real and reachable. A working, published number is a basic quality filter on a scraped list.

How to use it

Scrape phone numbers from acme.com.

Scrape phone numbers from each of these 40 domains. Return one row per domain
with every number found and label each as switchboard, sales, or unknown.

For each Google Maps listing with a website, scrape phone numbers from that
site and compare them to the number on the listing.

For the 10 accounts I flagged, find the mobile number for the named contact
from their LinkedIn URL instead.

Example prompts

Best prompting practices

• Set the expectation up front. Tell your client these are company lines, so it labels them as switchboard rather than dropping them into a mobile column.
• Pass the bare domain. You do not need a specific page — the crawl covers contact and location pages already.
• Ask for every number, tagged. “Return all numbers found with the page they came from, and mark which look like mobiles.” A single “main number” answer throws away the direct lines.
• Normalise the output. “Return numbers in international format with the country code” — otherwise a local-format number is useless to a dialler and cannot be passed to validate_whatsapp.
• Forbid guessing. Numbers must come from the page. A plausible-looking fabricated number wastes a dial and burns the account.
• Filter before you crawl. Rows with no website cost 0.5 credits to learn nothing. Drop them first and report the count.
• Escalate deliberately. find_work_phone and find_mobile_number cost 12 credits each. Reserve them for accounts you have already qualified.
• Validate mobiles separately. If a scraped number does turn out to be a mobile, validate_whatsapp tells you whether it is reachable there.

Output

{
  "url": "https://acme.com",
  "phones": ["+1 512-555-0142", "+1 800-555-0110"]
}

Credits and limits

0.5 credits per call. Run check_credits (free) before a large batch.

Parameters

Related tools

Next steps

• All MCP tools — the full catalogue with credit costs
• Research & Web Scraping — workflows that chain this tool into a list build

Keep learning

Two courses take these tools past the reference page — clustered by GTM job, then chained into workflows.