Tech Stack Detector: BuiltWith & Wappalyzer Alternative

Give it a list of websites and get back what each one runs (CMS, ecommerce platform, analytics, marketing and chat tools, payments, CDN, hosting) with evidence for every detection. Each row also carries what a homepage cannot show: the company's email provider, the services that send its email and the SaaS tools it has verified its domain with, read from public DNS, plus how the site treats AI crawlers. Pay per website read. Blocked and unreachable sites are free.

Built for sales and lead-generation teams who segment prospects by stack ("Shopify stores using Klaviyo", "Microsoft 365 companies that use Zendesk"), agencies qualifying leads, market and competitor research, and AI agents that need to enrich a domain in one call.

What you get for each website

Field group	Read from	Examples
Technologies	homepage headers, cookies, meta tags, script URLs, inline loader snippets, HTML, DNS (NS, CNAME)	WordPress 6.6, Shopify, Next.js 15, Google Tag Manager, HubSpot, Intercom, Stripe, Cloudflare, Vercel
Ready-made columns	the technologies above, grouped (frameworks such as Next.js go in js_frameworks, not cms)	cms, ecommerce, js_frameworks, analytics, tag_managers, advertising, marketing_automation, crm, live_chat, payments, cdn, hosting, web_servers, consent, ab_testing, reviews, security, ai_tools
Email	DNS MX, SPF, DMARC	provider (Google Workspace, Microsoft 365, ...), security gateway (Proofpoint, Mimecast, ...), sending services (Salesforce, Zendesk, HubSpot, SendGrid, Mailchimp, ...), DMARC policy
Verified SaaS	DNS TXT verification records	Atlassian, Slack, Zoom, Notion, Figma, DocuSign, Stripe, OpenAI, Anthropic, Cursor and 150 more vendors
AI-crawler policy	robots.txt, llms.txt, /.well-known/tdmrep.json, robots meta	verdict for 18 named AI bots (GPTBot, ClaudeBot, Google-Extended, PerplexityBot, ...) grouped as training, search and user-triggered; how many of 200+ known AI agents are blocked; Cloudflare Content-Signal; IETF Content-Usage; llms.txt; TDM reservation; noai
Evidence	per technology	confidence 0 to 100 and up to three short reasons, such as script https://js.stripe.com/v3/ or DNS NS ns-12.awsdns-01.com

Every input produces exactly one row with status = ok (homepage read), blocked (the site refused the visit), not_found (no such website) or error (timeout, server error, invalid input, or not processed because your maximum total charge was reached). example.com and www.example.com count as one website. A run never "succeeds" with an empty dataset: if nothing could be read, the run fails and says why.

Why this one

• DNS signals that page-only tools miss. In a sample of 200 well-known UK and US company domains read on 1 Nov 2026, 70% had a Microsoft 365 verification record, 59% Atlassian, 41% Anthropic, 38% OpenAI and 22% Cursor. None of that is visible on a homepage.
• Evidence and a confidence for every detection. Weak hints (a domain that is only allowed in a Content-Security-Policy header) score 40 and stay hidden at the default threshold of 50; set minConfidence to 0 to see them.
• Honest about blocks. When a site refuses the visit, the row says who refused it (Cloudflare, Akamai, Imperva, DataDome and others), keeps the DNS and robots.txt findings, and costs nothing.
• About 7,700 technology fingerprints, plus our own rules for chat and marketing tools whose standard install snippet only names its loader URL (Intercom, Drift, Microsoft Clarity, Mixpanel, Pardot, Gorgias and others).
• Fair pricing. $3 per 1,000 websites read. Blocked, not-found and error rows are free, there is no minimum run cost, and the DNS and AI-crawler checks are included.
• Agent-ready. List input, typed JSON output with a documented schema, ten websites in parallel by default, works from the Apify MCP server.

Input

{
  "domains": ["octopus.energy", "https://www.thortful.com/", "jane@pukkaherbs.com"],
  "checkDns": true,
  "checkAiPolicy": true,
  "includeEvidence": true,
  "minConfidence": 50
}

• domains: domains, URLs or email addresses (their domain is used). Duplicates are removed.
• checkDns: email provider, email senders, DMARC and verified SaaS from public DNS.
• checkAiPolicy: robots.txt verdicts for AI crawlers, llms.txt, TDM reservation and noai tags.
• includeEvidence: up to three reasons per technology; switch off for smaller rows.
• minConfidence: 0 to 100, default 50.
• respectRobotsTxt (default off): skip homepages whose robots.txt disallows all visitors; those rows are free.
• maxConcurrency: websites checked at the same time (1 to 50, default 10).

Output example

One real row, read from octopus.energy on 1 Nov 2026 and shortened (the full row lists 11 technologies and 22 verified services). The email provider sits behind a Mimecast gateway, so the MX records alone would only show Mimecast; the SPF record names Google Workspace.

{
  "status": "ok",
  "domain": "octopus.energy",
  "final_url": "https://octopus.energy/",
  "title": "Octopus Energy: The UK's most awarded energy supplier",
  "js_frameworks": ["Next.js", "React"],
  "hosting": ["Amazon Route 53", "Amazon Web Services", "Vercel"],
  "reviews": ["Trustpilot"],
  "email_provider": "Google Workspace",
  "email_security_gateway": "Mimecast",
  "email_senders": ["SendGrid", "Google Workspace", "Mimecast", "HubSpot"],
  "dmarc_policy": "reject",
  "verified_service_names": ["Microsoft 365", "Zoom", "Anthropic", "Atlassian", "Figma", "HubSpot", "OpenAI", "Slack", "Stripe"],
  "ai_policy_summary": "allows_ai",
  "llms_txt": false,
  "email": {
    "provider": "Google Workspace",
    "provider_evidence": "SPF include:_spf.google.com (MX is Mimecast)",
    "mx_hosts": ["eu-smtp-inbound-1.mimecast.com", "eu-smtp-inbound-2.mimecast.com"]
  },
  "technologies": [
    {"name": "Next.js", "categories": ["JavaScript frameworks", "Web frameworks", "Web servers", "Static site generator"],
     "version": null, "confidence": 100, "evidence": ["script https://octopus.energy/_next/static/immutable/chunks/26827qemr2pt4.js"]},
    {"name": "Vercel", "categories": ["PaaS"], "version": null, "confidence": 100,
     "evidence": ["header server: Vercel", "header x-vercel-cache: HIT"]},
    {"name": "Trustpilot", "categories": ["Reviews"], "version": null, "confidence": 75,
     "evidence": ["inline script loads //widget.trustpilot.com/bootstrap/v5/tp.widget.bootstrap.min.js"]}
  ]
}

The dataset has four table views: Tech stack, Email & SaaS (DNS), AI-crawler policy and All technologies. Export as JSON, CSV or Excel, or read it through the API.

Pricing

Event	Price	When
Website	$0.003 ($3 per 1,000)	the homepage was read and analysed
Actor start	$0.00005 per GB of run memory	once per run: $0.0002 at the default 4 GB

Blocked, not-found and error rows are not charged. Set a maximum total charge on the run and the actor never goes over it: each website reserves its charge before it starts, and inputs beyond the limit still get a row (status error, category budget) at no cost.

Limits

• HTTP only, homepage only. Technologies that show themselves only through JavaScript running in a browser (about a fifth of the fingerprint set) are missed unless their install snippet names a loader URL we recognise. Inner pages are not visited.
• Some sites refuse automated visits. Those rows are marked blocked, name the protection when it can be recognised, still carry DNS and robots.txt findings, and are free. No proxies or CAPTCHA tricks are used.
• DNS signals are read for the domain you give (without www.). For a subdomain such as shop.example.com, also give the company's main domain to get its email and SaaS signals; platform domains (myshopify.com, github.io, ...) are never used as a stand-in.
• DNS lookups use the platform's resolver first. When it cannot answer (some networks drop large answers), the same question goes to public resolvers (Cloudflare 1.1.1.1, Google 8.8.8.8). A lookup that still fails leaves the DNS fields empty; it never reads as "no records".
• A DNS verification record shows that the organisation set up an account with a vendor at some point. It does not prove the tool is in use today.
• Unusual pages are handled defensively. Fingerprint checks have time limits; when one is cut short the row carries analysis_limited: true. When a site's HTTPS certificate fails, the page is read without verification (read-only) and tls_valid is false. Domains that point to private or internal addresses are not fetched.
• The AI-crawler policy reports what the site states (robots.txt, llms.txt, TDMRep, tags). It does not measure whether crawlers comply. When robots.txt cannot be fetched, the summary is unknown.

Data sources and licences

Technology detection uses the community fingerprint set enthec/webappanalyzer (GPL-3.0), run here as a hosted service. AI crawler names come from the ai.robots.txt project (MIT). Bot purposes come from each vendor's own documentation. This actor is not affiliated with BuiltWith or Wappalyzer.

Support

Open an issue on this page; issues are answered within 12 hours on working days. Feature requests are welcome: a "changes since last run" monitor (technology added or removed, new verified SaaS) and inner-page scanning are on the list.