Hosting Provider Detector | Host & CDN Signals

Inspect hosting signals across a domain list

Hosting Provider Detector: Find hosting, PaaS, CDN and web-server signals exposed in public HTML and headers. Export infrastructure context for website research. An edge provider may hide the origin; this is not authoritative DNS or ownership data.

Start with a small example

1. Open Input, switch to JSON and paste the example below. Replace the example startUrls with your own research targets when ready.
2. Check the live Pricing tab and set a run spending limit.
3. Run the Actor, then open the Workflow output view. Inspect url, hosting, paas, cdn, webServer, error.
4. Export JSON for nested data or CSV for a spreadsheet. Review a small sample before increasing the scope.

{
  "startUrls": [
    "https://vercel.com",
    "https://wpengine.com",
    "https://kinsta.com"
  ],
  "filterTechnologies": [
    "Amazon Web Services",
    "Google Cloud",
    "Microsoft Azure",
    "Cloudflare",
    "Vercel",
    "Netlify",
    "WP Engine",
    "Kinsta",
    "SiteGround",
    "Fly.io",
    "Render",
    "GitHub Pages",
    "Heroku",
    "DigitalOcean",
    "Pantheon",
    "Flywheel",
    "GoDaddy Website Builder",
    "Bluehost",
    "Hostinger",
    "OVH",
    "Hetzner",
    "Linode",
    "Akamai",
    "Fastly",
    "Amazon CloudFront",
    "Google Cloud CDN",
    "Firebase",
    "Amazon S3",
    "Windows Server",
    "IIS",
    "Nginx",
    "Apache HTTP Server",
    "LiteSpeed",
    "Caddy",
    "OpenResty"
  ],
  "onlyMatching": false
}

At the rate checked September 9, 2026, 3 saved rows cost $0.009 in Actor event fees ($3.00 per 1,000 rows). Actual returned rows determine the event fee. Taxes and other account services may be separate.

What can I use it for?

• Hosting and managed-WordPress sales. Build a list of sites on a competitor's platform, or of sites on shared hosting that would benefit from managed infrastructure.
• Agency and MSP prospecting. Sites on GitHub Pages, Netlify or Vercel are run by developers; sites on Bluehost or GoDaddy are usually run by their owners. Target accordingly.
• Infrastructure audits. Inventory which providers, CDNs and servers a portfolio of client or subsidiary sites actually runs on.
• Competitive research. See what the fast, well-run sites in your category are hosted on.
• Market sizing. Measure the adoption of Vercel, Netlify, WP Engine or Cloudflare across an industry or region.
• Enriching your data. Add hosting, cdn and webServer columns to a spreadsheet of domains in one run.

What data does it return?

One record per website:

Field	Meaning
url, finalUrl, statusCode, title	What you asked for, where it resolved to, and the page title
hosting	Hosting provider signals, e.g. Amazon Web Services, WP Engine, Kinsta
paas	Platform signals, e.g. Vercel, Netlify, Heroku, Fly.io, Render
cdn	Content delivery network, e.g. Cloudflare, Akamai, Fastly, Amazon CloudFront
webServer	Nginx, Apache HTTP Server, IIS, LiteSpeed, Caddy and others
reverseProxy	Edge and proxy layers where exposed
matchesFilter	true when the site uses any of the providers you listed under Providers to look for
matchedTechnologies	Which listed providers were found
cms, ecommerce, javascriptFrameworks	The rest of the stack, for context
technologyCount	Total technologies detected
technologies	The complete list with categories, versions and confidence scores
error	null on success, otherwise why the site could not be checked

Sites that time out or refuse the connection are still returned with an error, so input and output line up one to one.

Illustrative output

Values explain the output shape; they are not a live result or guaranteed field coverage.

{
  "url": "https://vercel.com",
  "finalUrl": "https://vercel.com/",
  "statusCode": 200,
  "title": "Vercel: Build and deploy the best web experiences with the AI Cloud",
  "hosting": "Amazon Web Services",
  "paas": "Vercel",
  "cdn": null,
  "webServer": null,
  "matchesFilter": true,
  "matchedTechnologies": ["Amazon Web Services", "Vercel"],
  "javascriptFrameworks": "Next.js, React",
  "technologyCount": 7,
  "technologies": [
    { "name": "Vercel", "categories": ["PaaS"], "version": null, "confidence": 100, "website": "https://vercel.com" },
    { "name": "Amazon Web Services", "categories": ["Hosting"], "version": null, "confidence": 100, "website": "https://aws.amazon.com/" }
  ],
  "error": null
}

How to use it

1. Paste your domains under Websites to check, upload a file, or pass them through the API as startUrls. Bare domains like example.com are fine.
2. Leave Providers to look for at its defaults for a broad inventory, or narrow it to one provider to build a targeted list.
3. Switch on Store only matching websites to pay only for sites on the providers you listed and discard the rest.
4. Click Start and download the results as JSON, CSV or Excel from the Output tab, or read them from the dataset via the API.

Input example

{
  "startUrls": [
    "https://vercel.com",
    "wpengine.com",
    "https://kinsta.com"
  ],
  "filterTechnologies": ["WP Engine", "Kinsta", "Vercel"],
  "onlyMatching": false
}

Raise Max concurrency for large lists. Turn on Proxy configuration only if some sites block the platform's IP range — most do not.

How detection works

Hosting and CDN providers announce themselves in response headers and asset URLs: x-vercel-id, server: netlify, x-powered-by: WP Engine, x-kinsta-cache, via: 1.1 google, x-amz-cf-id, cf-ray, x-akamai-transformed, x-served-by: cache-… for Fastly, x-github-request-id for GitHub Pages, and dozens more. The detector fetches each homepage once and checks the headers, cookies, HTML, script sources and meta tags against more than 7,600 technology fingerprints, then reports every match with a confidence score.

Because the page's JavaScript is never executed, signals that only appear after scripts run are not evaluated. For hosting that rarely matters — the headers on the first response are the evidence — and the trade-off buys speed and a much lower cost per site.

The fingerprint database is the open-source webappanalyzer technology dataset, used under the GNU GPL v3. The matching engine in this actor is original code.

Pricing

You pay per website saved to the dataset. With Store only matching websites on, that means per matching site; with it off, per site checked. There are no subscriptions, seats or minimums, and you can cap the maximum spend of a run before it starts.

Scope and responsible use

The actor requests only the publicly served homepage of each URL you provide, exactly as a browser would. It requires no login, no cookies and no credentials, performs no DNS enumeration, port scanning or IP-range lookups, does not crawl beyond the page you give it, and does not attempt to bypass any access control or bot-protection challenge. You are responsible for using the results in accordance with the laws that apply to you and the terms of the websites you check.

Limitations

• A site behind Cloudflare or another CDN can hide its origin host entirely; in that case the CDN is reported and hosting may be empty.
• Providers that strip identifying headers are reported only when other signals remain.
• Very large retail and media homepages that deliberately stall automated traffic can exceed the request timeout; they come back as rows with an error rather than being dropped.
• Only the URL you supply is fetched; a site whose marketing pages and application run on different infrastructure is reported for the page you gave.

Something not detected?

Open an issue on this actor's Issues tab with the URL and the provider you expected. Include a reproducible input and the expected signal so the report can be investigated.

Interpret empty results and errors

A null technology field means no recognized signal was found in the inspected response; it does not prove absence. A returned error means that URL was not successfully analyzed. With onlyMatching: true, non-matches and errors are omitted from the dataset, so a zero-row result cannot distinguish an unavailable site from no matches. Start with onlyMatching: false when checking coverage.

Every saved row, including an error row, incurs the configured per-row event fee. Proxy selection and run duration can affect operating costs. Do not increase concurrency or enable a more expensive proxy merely to work around an unclear result.

For a reproducible problem, open this Actor’s Issues tab with a small public input, expected behavior and relevant error text. Remove tokens and confidential information. No response-time or uptime guarantee is offered.

Use the result in an automation

In Make or n8n, use the Apify integration to run this Actor with the same JSON input, wait for completion, then retrieve its default dataset. Route failed runs and error rows to a review step before sending valid results to your spreadsheet or CRM. Scheduling does not make these Actors emit only new records: deduplicate downstream using the source URL or record ID.

For Node.js, install the official apify-client package and set your own APIFY_TOKEN environment variable. The following example starts a paid run with a small spending limit.

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const input = {
  "startUrls": [
    "https://vercel.com",
    "https://wpengine.com",
    "https://kinsta.com"
  ],
  "filterTechnologies": [
    "Amazon Web Services",
    "Google Cloud",
    "Microsoft Azure",
    "Cloudflare",
    "Vercel",
    "Netlify",
    "WP Engine",
    "Kinsta",
    "SiteGround",
    "Fly.io",
    "Render",
    "GitHub Pages",
    "Heroku",
    "DigitalOcean",
    "Pantheon",
    "Flywheel",
    "GoDaddy Website Builder",
    "Bluehost",
    "Hostinger",
    "OVH",
    "Hetzner",
    "Linode",
    "Akamai",
    "Fastly",
    "Amazon CloudFront",
    "Google Cloud CDN",
    "Firebase",
    "Amazon S3",
    "Windows Server",
    "IIS",
    "Nginx",
    "Apache HTTP Server",
    "LiteSpeed",
    "Caddy",
    "OpenResty"
  ],
  "onlyMatching": false
};
const run = await client.actor('keystonelabs/hosting-provider-detector').call(input, {
  memory: 512, timeout: 180, maxTotalChargeUsd: 0.10
});
if (run.status !== 'SUCCEEDED') throw new Error('Run did not succeed: ' + run.id);
const { items } = await client.dataset(run.defaultDatasetId).listItems({ limit: 100 });
console.log(items);

The first 100 rows are retrieved in this example. Use the dataset API pagination for a larger result. Keep credentials out of shared inputs and source files.

A useful next step

• Website Tech Stack Detector | Bulk URL Lookup: qualify a website list by its technology. This is a separate Actor with its own input and price.

Understand the run at a glance

Open Run summary in Output for counts of unique URLs, analyzed responses, failures, matches and saved rows. It remains useful when onlyMatching produces an empty dataset. The summary is written at normal completion and does not add a billable dataset row. Failed or aborted runs may not have a final summary; check run status first. The summary contains counts only, not a second copy of scraped content.