Website Lead Extractor: Emails, Phones and Company Addresses

Give it a list of company websites. Get back one clean row per company, ready to paste into a CRM.

Most contact scrapers hand you a pile of page-level rows and leave you to stitch them together. This one returns a company record: the best email, the phone number in E.164, the postal address split into street, city and postcode, the VAT and company registration numbers, and the social profiles. Every email is checked against DNS before you see it.

You pay per company, not per page. If a site publishes nothing, you are not charged.

---

What you get for every website

Field	Example	Notes
bestEmail	sales@northwind.co	The single address most likely to reach a human
emails[]	6 addresses	Each with department, role flag and deliverability
bestPhone	+442079460958	E.164, ready for a dialler or a CRM
phones[]	3 numbers	With national format and line type
street, city, region, postalCode, countryCode	12 Baker Street, London, NW1 6XE, GB	Parsed, not a blob of text
addresses[]	every office found	Each with a confidence level and its source page
poBoxes[]	PO Box 4321	Recognised in 8 languages
vatNumber	GB123456789	Validated against the country's real VAT format
companyNumber	08123456, HRB 6089	UK, German, Dutch, French and Nordic registries
linkedin, twitter, facebook, instagram, youtube, tiktok	canonical profile URLs	Share buttons and tracking links removed
mxProvider	Google Workspace	Who runs their mail, from the MX records
companyName	Northwind Ltd	From the site's own declared name
found	true	false means nothing was published, and nothing was charged
error, errorDetail	blocked, plus what to do about it	Plain English, not a stack trace

Export as JSON, JSONL, CSV or Excel, or pull it straight from the API.

---

How to use it

Paste domains into Websites to extract leads from. Any of these work, and all three of these are recognised as the same company and charged once:

northwind.co
https://www.northwind.co/contact
NORTHWIND.CO/

You can also paste a whole spreadsheet column separated by commas or new lines. That is it. Everything else has a sensible default.

Input options

Option	Default	What it does
startUrls	required	Websites to enrich. Domains or full URLs
maxPagesPerDomain	5	Pages read per site. Costs you nothing extra, because billing is per company
verifyEmails	true	MX lookup, disposable detection, mail provider detection
enableBrowserFallback	true	Retries with a real browser when a site is JavaScript-only or blocks plain requests
maxWebsites	unlimited	Hard spend cap on large inputs
maxConcurrency	10	Sites processed at once
proxyConfiguration	datacenter	Switch to Residential for sites behind heavy bot protection

---

Why the results are cleaner

Most of the work in this Actor is not finding contact details. It is refusing to return things that only look like contact details.

Invented emails are the worst defect a lead list can have. They pass a visual check, they bounce, and the bounces damage your sending domain. Ordinary English is full of phrases like "reach them at northwind.co", and a naive scraper turns that into them@northwind.co. This one requires an explicit obfuscation marker, so sales [at] northwind [dot] co is decoded and "located at northwind.co" is left alone. Retina image filenames (logo@2x.png), Sentry keys and you@example.com placeholders are filtered out too.

Cloudflare-obfuscated addresses are decoded. Cloudflare's email protection XOR-encodes addresses so that regex scrapers see nothing. It is on by default for a large share of small business sites. Those addresses are decoded and returned.

Third-party vendor emails are dropped. Scraping a shop that uses Intercom should not return Intercom's legal address. Around fifty common vendor domains are excluded unless you are scraping that vendor.

Share buttons are not social profiles. facebook.com/sharer, twitter.com/intent/tweet and LinkedIn's share widget are excluded. What you get is the company's actual page.

A date is not an address. "June 09, 2025" contains a street-shaped fragment and a four digit number that is a valid postcode in Australia. Dates, years, IBANs, court registrations and cookie notices are all rejected before an address is accepted.

Every address carries its confidence. high means the site published structured schema.org data and no guessing was involved. medium is a semantic <address> element. low is a scored read of the footer or contact page, and it always includes the untouched original text so you can check it.

---

Where the contact details come from

The Actor reads the homepage, then picks the few pages that actually carry contact details: contact, about, team, locations, and the legal notice. It does not crawl blog posts or product listings.

Legal notice pages are prioritised almost as highly as contact pages. In Germany, Austria and Switzerland the Impressum is legally required to carry the full postal address, the managing director and the VAT identification number, which makes it the richest page on most European sites.

Five pages is enough for nearly every website. Raising the limit finds a little more on very large sites and costs you nothing extra.

---

Pricing

$10 per 1,000 companies enriched. One charge per company that produced at least one contact.

You are not charged when:

• the domain does not exist
• the site cannot be reached
• the site is reachable but publishes no contact details at all
• the only thing found was a generic Instagram or Facebook link

Compare that to paying per page. A crawler at $1.05 per 1,000 pages reading 20 pages per site works out to about $21 per 1,000 companies, and gives you no addresses, no VAT numbers, no E.164 phones and no email verification.

---

Common uses

Building a B2B outreach list. Export a domain column from Apollo, Crunchbase or a directory, run it through here, and get verified emails with the role and department already labelled so you can route sales enquiries differently from support enquiries.

Enriching a CRM that has domains but no contacts. One row per company maps directly onto a company record. bestEmail and bestPhone are single values, so there is nothing to reduce or deduplicate on import.

Supplier and counterparty due diligence. The registered address, VAT number and company registration number are what a KYB or onboarding check actually needs, and they come from the company's own legal notice page.

Direct mail and field sales. Postal addresses are parsed into street, city, region, postcode and ISO country code, so they are ready for a mailing house without manual cleanup.

Cleaning a list you already have. Run it with verifyEmails on and use the deliverability field to drop addresses that would bounce before you load them into a sending tool.

Local business research. Phone numbers come back in E.164 with the line type, so mobile numbers can be separated from switchboards.

---

Email verification, described honestly

Verification is done entirely with DNS. For each address the Actor checks the syntax, looks up the domain's MX records, checks a bundled list of over 8,000 disposable email domains, and identifies the mail provider.

deliverability	Meaning
valid	The domain has MX records and is not disposable. Mail can be delivered to it
risky	Deliverable but worth a second look: a free mailbox provider, a disposable domain, or a domain with no MX record
invalid	Malformed, or the domain cannot receive mail at all
unknown	DNS did not answer. Never silently upgraded to valid

There is no SMTP probe, deliberately. Connecting to someone else's mail server to test whether a specific mailbox exists gets shared IPs blocklisted, is rate limited into uselessness at any real volume, and is treated as address harvesting by the receiving side. So this will tell you that a domain can receive mail and that an address is not a throwaway. It will not tell you that one particular mailbox exists. Any tool that claims otherwise at scale is either guessing or getting blocked.

---

Sites that block scrapers

Some commercial sites answer a plain HTTP request with a 403 and answer a real browser with the page. When that happens the Actor automatically retries in a headless browser, which clears most of them.

For the stubborn ones, set Proxy configuration to Residential and re-run just those domains. The errorDetail field tells you which rows to retry and why. Blocked rows are never charged.

---

Output example

{
  "domain": "northwind.co",
  "companyName": "Northwind Ltd",
  "found": true,
  "bestEmail": "sales@northwind.co",
  "bestPhone": "+442079460958",
  "bestAddress": "12 Baker Street, London, NW1 6XE, United Kingdom",
  "street": "12 Baker Street",
  "city": "London",
  "postalCode": "NW1 6XE",
  "countryCode": "GB",
  "vatNumber": "GB123456789",
  "companyNumber": "08123456",
  "mxProvider": "Google Workspace",
  "linkedin": "https://linkedin.com/company/northwind",
  "emails": [
    {
      "email": "sales@northwind.co",
      "department": "sales",
      "isRoleAccount": true,
      "isSameDomain": true,
      "isFreeProvider": false,
      "validation": {
        "deliverability": "valid",
        "reason": "mx_record_present",
        "hasMx": true,
        "mxProvider": "Google Workspace"
      },
      "sourceUrl": "https://northwind.co/contact"
    }
  ],
  "phones": [
    {
      "e164": "+442079460958",
      "national": "020 7946 0958",
      "countryCode": "GB",
      "type": "fixed_line",
      "sourceUrl": "https://northwind.co/contact"
    }
  ],
  "pagesCrawled": ["https://northwind.co/", "https://northwind.co/contact"],
  "error": null
}

---

Frequently asked questions

How do I extract emails from a list of websites? Paste the domains into the input field and run it. One row comes back per company, with every email found on the site, each labelled by department and checked against DNS.

Can it scrape a company's phone number and address as well as the email? Yes. Phone numbers are returned in E.164 with the line type, and postal addresses are split into street, city, region, postcode and ISO country code.

Does it find the VAT number or company registration number? Yes, when the site publishes them, which most European companies must. UK company numbers, German HRB numbers, Dutch KvK numbers, French SIREN and SIRET numbers and EU VAT identifiers are recognised, and each is validated against its country's real format so that an order reference is never returned as a VAT number.

What happens if a website has no contact details? You get a row with found: false and empty arrays, and you are not charged for it.

How many pages does it read per website? Five by default, chosen by relevance rather than crawled breadth first. You can raise it to 25. It does not change the price.

Is this GDPR compliant? The Actor reads only what a company publishes on its own public website, which is typically business contact data rather than personal data. How you use the output is your responsibility: if you contact people in the EU or UK, you need a lawful basis and you must honour opt-outs. This tool does not extract named individuals' personal profiles.

Can I run it on a single website? Yes. One domain in, one enriched company record out, for one cent.

How is this different from a generic contact details scraper? Three things: one row per company instead of one per page, postal addresses and registration identifiers that generic scrapers do not return at all, and DNS verification on every email. Plus per-company pricing, which works out at less than half the effective cost of per-page crawling.

---

Other tools by the same author

• LLM-Ready Web Extractor: any URL to clean Markdown or JSON
• SEO Audit and Site Health Checker: on-page SEO issues and a health score per page
• Link Preview and URL Metadata API: Open Graph and metadata extraction, free
• AgentReader: clean web content for AI agents, over MCP

---

Notes

Public data only. The Actor reads pages that any visitor can load, and it does not log in, bypass paywalls or collect data from behind authentication.