Company Website & Profile Resolver (with evidence)

Give an agent or a CRM a company name or a domain. Get back the official website and the public pages that site actually links to, each claim tied to a URL, a short snippet, and the time it was observed. A row is billed only when the evidence score meets your threshold.

Unofficial. Not affiliated with LinkedIn, Brave Search, Serper, Greenhouse, Lever, Ashby, Apple, Google, or any company you resolve.

Who it is for

AI agents and enrichment pipelines that must turn "Acme" into the right domain without inventing one. RevOps and CRM cleanup, when a spreadsheet has legal names and you need the real site, careers page, and docs. Scheduled watchers that should alert only when the resolved identity changes.

It does not guess facts with a language model. If a page does not say it, the field stays empty.

What you get

For each company:

• Canonical official website, after redirects
• Social profile URLs that the site itself publishes (LinkedIn company page included only in that case)
• Careers page and, when the HTML exposes one, the ATS vendor and board slug
• Pricing, docs, contact, and status-page links
• App-store or marketplace listings linked from the site
• Sitemap size from the first sitemap document
• Confidence from 0 to 1, floored to two decimals
• evidence[] with claim, url, snippet, snippetSha256, observedAt
• status: resolved, low_confidence, not_found, blocked, no_results, or unsupported

A run summary is stored in the default key-value store under OUTPUT (rowsCharged, rowsDropped, fillRate, coverage, and error counts).

Example

Trimmed from a local run of the default input on 23 October 2026. charged is true here because the pay-per-event test path accepted the row. On Apify it is true only after company-resolved is billed.

{
  "input": "Stripe",
  "status": "resolved",
  "officialWebsite": "https://stripe.com/",
  "companyNameObserved": "Stripe",
  "confidence": 1.0,
  "confidenceThreshold": 0.8,
  "charged": true,
  "chargeEvent": "company-resolved",
  "coverage": "partial",
  "coverageNote": "No search API key was used. Candidates came from the supplied domain or URL and conservative domain guesses.",
  "linkedinCompanyUrl": "https://www.linkedin.com/company/stripe",
  "githubUrl": "https://github.com/stripe",
  "careersUrl": "https://stripe.com/careers",
  "pricingUrl": "https://stripe.com/pricing",
  "docsUrl": "https://docs.stripe.com/",
  "contactUrl": "https://stripe.com/contact/sales",
  "statusPageUrl": "https://status.stripe.com",
  "sitemapUrl": "https://stripe.com/sitemap/sitemap.xml",
  "sitemapSize": 9,
  "sitemapKind": "index",
  "fillRate": 1.0,
  "evidence": [
    {
      "claim": "registrable domain matches the company name",
      "url": "https://stripe.com/",
      "snippet": "stripe.com",
      "snippetSha256": "1ffda32ad70e3cac6d6a04e2fbb00c062cc12d3a02796149e639c6d9a3e2774b",
      "observedAt": "2026-10-23T14:11:55Z"
    },
    {
      "claim": "homepage title matches the company name",
      "url": "https://stripe.com/",
      "snippet": "Stripe | Financial Infrastructure to Grow Your Revenue",
      "snippetSha256": "9b75a03d7f02b799048b25ef8699bfdce88e231ea5a699b189a8a01dda2604a8",
      "observedAt": "2026-10-23T14:11:55Z"
    },
    {
      "claim": "schema.org Organization name matches the company",
      "url": "https://stripe.com/",
      "snippet": "Stripe",
      "snippetSha256": "961a6062ba1ba2d8608e5eea8be67483d936f2f9f41e312217407198e14f087c",
      "observedAt": "2026-10-23T14:11:55Z"
    },
    {
      "claim": "footer or schema.org legalName names the company",
      "url": "https://stripe.com/",
      "snippet": "Stripe, LLC",
      "snippetSha256": "26a7c57c0c19bec1a40b7b16570c2b8b488006e10e4c6b6173905af27c4d5c83",
      "observedAt": "2026-10-23T14:11:55Z"
    }
  ]
}

sitemapKind: index means 9 child sitemaps were listed. Those children are not downloaded, so this is not a count of every URL on the site.

How candidates are chosen

With a search API key, the Actor queries Brave Search or Serper, drops directories and social hosts (Wikipedia, LinkedIn, GitHub, news sites, and similar), and fetches the remaining candidates plus a conservative guess.

Without a key, coverage is partial and the row says so. Candidates are only:

• the domain or URL you supplied, and
• one or two guesses: the company slug on .com, and, when you set a country hint, that country's TLD.

No other Apify Actor is called. LinkedIn is never fetched. A LinkedIn company URL is copied only when the company site publishes it. Personal /in/ profiles are discarded. Email addresses are redacted and not returned.

Confidence

The score is a weighted sum of what the fetched HTML contained, then floored to two decimals. It is not a model probability.

Signal	Weight
Domain matches the company slug, or the page is the domain you supplied	0.32
Title, site name, or schema.org name matches	0.26
schema.org Organization or legal name matches	0.16
Footer copyright or schema.org legalName matches	0.10
Contact link on the same site	0.06
Social profile linked from the site	0.06
Careers, pricing, docs, or status link	0.04

Hard caps keep weak pages below the default 0.8 charge line:

• Parked domain or a directory host scores 0 and is not the official site.
• schema.org names a different organization: at most 0.30.
• No on-page name match: at most 0.35.
• Name matches but the domain was not supplied and does not match: at most 0.40.
• Article or news page without an Organization match: at most 0.55.
• Country or city hint not visible in the title, footer, or schema.org address, and the TLD does not match: at most 0.74. A US hint does not apply this cap.
• Fewer than three independent signals: at most 0.79.
• Two different domains both score near the threshold: the row stays low_confidence instead of picking one.

A GitHub profile linked from the site can add 0.04 when that profile links back and the score is just under the threshold. That is the only extra fetch of a social site, and it is skipped when the homepage already clears the threshold.

Pricing

You're only charged for rows that pass validation. company-resolved fires only when status is resolved, confidence is at least your threshold, the website is a public https URL, there are at least two complete evidence items, and the fill rate is at least 0.5. low_confidence, not_found, blocked, no_results, and unsupported are free. The same input is not charged twice in one run.

There is no per-dataset-item charge. Prices:

Event	Price	When it fires
company-resolved	$0.02	A validated row at or above the threshold
unresolved rows	$0	Anything that does not pass the rule above
apify-actor-start	$0.005 per GB (platform start event)	Once per run

At the default 512 MB, the start event is about $0.0025 per run.

Run	Billable rows	Event total
Default demo, 8 companies, all resolved	8	8 × $0.02 = $0.16
100 companies, 70 resolved, 30 not found or low confidence	70	70 × $0.02 = $1.40
1,000 resolved companies	1,000	$20.00

maxCompanies defaults to 25 (about $0.50 if every row resolves). maxChargeUsd defaults to $5 (about 250 resolved rows). The run also stops when Apify reports the platform max total charge. Rows already delivered stay. A resolved company that would exceed the cap is not returned as a free success.

Input

Every load-bearing field has a default. An empty {} run uses the demo list: Stripe, Notion, Figma, Vercel, Shopify, Atlassian, HubSpot, Cloudflare.

Field	Default	What it does
companies	the 8 demo names	Names, domains, or URLs
companiesText	empty	Extra companies, one per line. Name, domain.com is accepted
countryHint	empty	Disambiguation only
cityHint	empty	Disambiguation only
confidenceThreshold	0.8	Charge line
searchProvider	none	none, brave, or serper
searchApiKey	empty, secret	Your Brave or Serper key. Never required
maxCompanies	25	Cap, 1 to 500
outputMode	all	changes_only emits new or changed identities
maxChargeUsd	5	Actor-side dollar cap. 0 disables it

Private IPs, localhost, file://, credentialed URLs, and ports other than 80 and 443 are rejected and returned as unsupported. They are not fetched.

Repeat runs

Resolved identities are stored in the named key-value store company-web-presence-state. A later blocked or not_found result does not overwrite a good baseline. changes_only pushes a dataset row only when the website, LinkedIn URL, careers URL, or ATS slug differs, or when the company has never been resolved. If nothing changed, the dataset gets one free no_results row so the run is not silently empty.

OUTPUT on every run includes counts, rowsCharged, rowsDropped, fillRate, chargeLimitReached, coverage, and errors (proxy, rateLimit, blocked, network, other). An HTTP 200 challenge page counts as blocked, not as a match.

Integrations

The dataset is flat enough for CSV (appStoreUrlsFlat joins store links with pipes). evidence stays an array.

• Webhook. Point an Apify webhook at the dataset. Filter on status == resolved or, for a watcher, run with outputMode: changes_only.
• n8n / Make. Use the Apify node, pass companies from a sheet or CRM, and map officialWebsite, linkedinCompanyUrl, careersUrl, and confidence.
• Google Sheets. Use the Apify integration to import the default dataset. Empty profile fields are null, not invented strings.
• MCP. The input and dataset schemas name every field. Call the Actor with {} and the demo set runs. Read OUTPUT for rowsCharged before you trust a batch.

Scope and limits

• Public HTML only. No login, no CAPTCHA solving, no browser. A site that is an empty JavaScript shell can come back low_confidence or not_found.
• Without a search key, a generic name can miss or, more rarely, land on a different company that owns the .com and uses the same name. Close scores are returned as low_confidence instead of a forced pick.
• Sitemap size is the first sitemap or sitemap index only.
• ATS slug is set only when a known board URL (Greenhouse, Lever, Ashby, Workable, and similar) appears in the homepage or careers HTML.
• This is not a legal registry check, a sanctions screen, or a guarantee of corporate identity.
• Polite fetching: an identified User-Agent, per-host delay, short timeouts, and one retry on 429 and 5xx. The contact address in the User-Agent is plusonecares@gmail.com.
• Page text is treated as untrusted. Snippets are short plain text. Emails are stripped. Nothing on the page is executed or followed as an instruction.
• Do not use the output to email people. This Actor does not collect personal contact data. Outreach still has to follow the rules that apply to you.

Source pages remain the property of their publishers. The Actor records short factual excerpts so you can audit a claim. It does not grant a licence to republish those sites.

Run locally

Python 3.12.

pip install -r requirements-dev.txt
pytest
RUN_LIVE=1 pytest tests/test_live_smoke.py
python -m src

python -m src reads Apify local storage input, or INPUT.json in the working directory. With neither, the demo list is used. Local runs set ACTOR_TEST_PAY_PER_EVENT so charge calls are logged and not billed.

On 23 October 2026 the default eight companies resolved 8/8 in about 7 seconds, with no search key: precision 1.00 and coverage 1.00 against those known domains. The fixture eval in tests/ (11 companies that should resolve, plus parked, article, directory, wrong-brand, ambiguous, and blocked cases) also scored precision 1.00 and coverage 1.00.

FAQ

Why was a row free? Confidence was under the threshold, the page was blocked or ambiguous, or the row failed the fill-rate check. reason says which. rowsDropped in OUTPUT counts rows that looked resolved but failed validation.

Do I need a search key? No. You will get coverage: partial and domain guesses only. Brave and Serper are optional bring-your-own-key providers.

Do you scrape LinkedIn? No. The company URL is copied from the site you resolved. The LinkedIn page is not requested.

Can I raise the batch size? Yes. Set maxCompanies and maxChargeUsd. The run stops cleanly at the cap.

What if two companies share a name? You get low_confidence and both domains in evidence, not a silent pick.

Questions: plusonecares@gmail.com.