Company Registry Search — Companies House, SIRENE +9 (KYB)

Verifying a counterparty in more than one country means a different official register, identifier format and response shape for each one, most of them answering in their own language. This Actor searches the official company registers of 11 countries by company name or registration number, in parallel, and returns one normalized record per matched entity — same field names whatever country it came from, plus the URL of the official register entry it was read from.

What this does

• Company registry search by name or by registration number. Pass a company name or any covered register's own primary identifier (see the Input table). Each country decides independently whether the query looks like one of its numbers; if not, it name-searches there — a nine-digit query is a French SIREN, a Norwegian orgnr and an Australian ACN at once, so per-register detection is the only form of the question with an answer.
• Pass a list, not just one name. queries takes up to 100 names and/or registration numbers, each fanned out over the same countries — one run for a whole supplier, vendor or portfolio list, with every row carrying the query it came from so the results split back apart.
• One normalized record per entity, whatever the source: legal name and registration number, status verbatim and mapped to active / inactive / unknown, legal form, incorporation date, registered address as one line plus components, activity codes carrying the source's own scheme name (NAF, NACE, CZ-NACE, PKD, TOL2008), officers where published, GLEIF LEI, and source_url — a citation on every record.
• Officers / directors for the three registers that publish them under an open licence: Companies House (officers page), Annuaire des Entreprises (dirigeants) and Brønnøysundregistrene (roller). Everywhere else officers is [] with an officers_note saying why, so a record never looks truncated.
• GLEIF LEI enrichment on every matched entity, joined on registration number and falling back to legal name plus country. A LEI is attached only on a confident single match, and enrichment never delays the registry record.
• Explained misses. Every country that produced no entity still returns a row with a machine-routable outcome and a plain-language message. None of them is charged, and the run never ends FAILED.
• No API key, no account, no signup. Every source is a keyless public official register, read over plain HTTP requests — no browser, no data broker in the path.
• KYB and due-diligence use cases: verify a cross-border supplier before onboarding; resolve a legal name and number to a LEI for sanctions-adjacent screening prep; check that a subsidiary named in a contract exists in its stated jurisdiction; enrich a vendor-master record with the registered entity and its official address.

Coverage

Country	Official register	Name search	Number lookup	Officers
GB	Companies House	yes	company number	yes
FR	Annuaire des Entreprises (SIRENE / INPI data)	yes	SIREN / SIRET	yes
NO	Brønnøysundregistrene — Enhetsregisteret	yes	organisasjonsnummer	yes
FI	Finnish Patent and Registration Office (PRH / YTJ)	yes	business ID	—
CH	Zefix — Swiss Central Business Name Index	yes	UID (CHE…)	—
EE	e-Business Register	yes	registry code	—
LV	Uzņēmumu reģistrs (Latvian Register of Enterprises)	yes	regcode	—
CZ	ARES — Czech Business Register	yes	IČO	—
PL	KRS — National Court Register	no — number only	KRS number	—
AU	Australian Business Register (ABN Lookup)	yes	ABN / ACN	—
SG	ACRA	yes	UEN	—
LT	Registrų centras	deferred	deferred	—

Plus the GLEIF Global LEI Index for LEI enrichment on every matched record.

Three notes the rows carry themselves: Zefix's open dataset lists currently-registered entities only, so a dissolved Swiss company will not appear; the ACRA dataset is a monthly snapshot; the Australian Business Register publishes ABN-level details, narrower than a full company record, and no incorporation date. Lithuania is recognised but not searchable in this version — its public search now sits behind an interactive browser challenge — and returns a free row saying so rather than being dropped from the coverage list.

Why one normalized schema over primary sources matters

The single-country lookup is not the hard part. The hard part is that a cross-border KYB question — is this company real, in this jurisdiction, under this name, with this number? — means integrating a different official source per country, each with its own identifier grammar, its own status vocabulary in its own language, and its own idea of what an address is: a separate thing to build, monitor and re-fix every time one of them moves.

The alternative most tooling reaches for is a commercial aggregator: one API, one schema, no per-country work. The cost is that you are no longer reading the register — you are reading somebody's copy of it, on their refresh cycle, with no way to show an auditor where a field came from. A diligence file wants a citation, not a vendor's assertion.

So this Actor reads the primary sources directly and pays the normalization cost once, so the caller never does. status is kept verbatim and mapped, because comparing entities across countries needs the mapping and defending a conclusion needs the original. And where a register genuinely doesn't publish something — a Swiss status field, an Australian incorporation date, officers outside three countries — the record says so in a note instead of returning a bare null and leaving you to guess whether the company or the pipeline was empty.

How it compares to other ways of getting this data

Approach	Countries per call	Reads the official register	Normalized across countries	Citation per record	Key / account needed
Single-country registry scraper	1	yes	n/a	sometimes	sometimes
Thin wrapper over 2–4 registry APIs	2–4	yes	partly	rarely	sometimes
Commercial company-data aggregator	many	no — a licensed copy	yes	no	yes, paid
This Actor	up to 12 + LEI	yes	yes, one schema	yes, every record	no

The bet: breadth under one schema, over primary sources, with citations is the thing that is actually missing. Any one of those three alone already exists.

Input

Field	Type	Required	Default	Description
query	string	one of query / queries	—	A company name (e.g. Carrefour) or an official registration number (e.g. 00445790, 652014051, 923609016, 0112038-9, CHE-105.909.036, 45274649, 0000028860, 88000014675, 196800306E). Numbers are matched exactly; names are matched as the register matches them.
queries	array of strings	one of query / queries	—	A list of names and/or registration numbers to look up in one run — a whole supplier or portfolio list in a single call. Up to 100 entries, 200 characters each. Give this, query, or both: they are merged, query first, and near-duplicates (case, accent, punctuation) are searched once. Each entry fans out over the same countries, and every row carries its own query.
countries	array of strings		all covered	ISO-3166 alpha-2 codes, e.g. ["GB","FR","NO"]. Empty searches every covered country. An unrecognised code returns one free not_covered row naming it and never fails the run.
maxRecordsPerCountry	integer		10	Upper bound on matched entities returned per country, 1–50. Each delivered entity record is billed once; rows reporting no match, no coverage, a failure or a rate limit are always free.
includeOfficers	boolean		true	Include officers for the three registers that publish them openly (GB, FR, NO). Only the name, role and appointment date the register itself publishes are returned — never dates of birth, phone numbers or email addresses.
includeLei	boolean		true	Enrich each matched entity with its GLEIF LEI when the registration number, or the legal name plus country, gives a confident match. No confident match leaves lei null.

Output

One dataset record per matched entity. Real record, from a name search scoped to Norway:

{
  "outcome": "match",
  "query": "Equinor",
  "country": "NO",
  "registry": "Brønnøysundregistrene — Enhetsregisteret (Norway)",
  "legal_name": "EQUINOR ASA",
  "registration_number": "923609016",
  "status": "Registrert",
  "status_normalized": "active",
  "legal_form": "Allmennaksjeselskap",
  "incorporation_date": "1972-09-18",
  "registered_address": "Forusbeen 50, 4035, STAVANGER, Norge",
  "address_components": {
    "address": "Forusbeen 50",
    "postal_code": "4035",
    "city": "STAVANGER",
    "municipality": "STAVANGER",
    "country": "Norge"
  },
  "activity_codes": [
    { "scheme": "NACE", "code": "06.100", "label": "Utvinning av råolje" },
    { "scheme": "NACE", "code": "06.200", "label": "Utvinning av naturgass" }
  ],
  "officers": [
    { "name": "Anders Opedal", "role": "Daglig leder" },
    { "name": "Jarle Kjell Roth", "role": "Styrets leder" }
  ],
  "officers_note": null,
  "lei": "OW6OFBNCKXC4US5C7523",
  "source_url": "https://virksomhet.brreg.no/nb/oppslag/enheter/923609016",
  "retrieved_at": "2026-09-18T09:41:02Z",
  "match_type": "name",
  "message": null
}

A country that produced nothing returns a row in the same schema, with outcome != match, message filled and every entity field null:

{
  "outcome": "not_covered",
  "query": "Equinor",
  "country": "PL",
  "registry": "KRS — Polish National Court Register",
  "legal_name": null,
  "message": "Poland's National Court Register (KRS) is searchable here by KRS number only: its open API exposes entity lookups by number, and its name-search front end is not reachable without an interactive browser session. Supply the 10-digit KRS number (for example 0000028860) to get the full Polish record. Nothing was charged for this row."
}

outcome is the field to branch on: match (the only billable kind), no_match, not_covered, failed, rate_limited, invalid_input, internal_error. Beside it, retry says what to do about a non-delivery in one value — later, with_changes, or never, and null on a match. The run also writes two key-value-store records — OUTPUT with the per-outcome counts, and COVERAGE with the live coverage manifest (which countries this build searches, which publish officers, each register's caveat) — so a caller can read current coverage instead of trusting a document.

Example

{
  "query": "Equinor",
  "countries": ["NO", "GB"],
  "maxRecordsPerCountry": 5
}

curl -X POST "https://api.apify.com/v2/acts/shelvick~global-company-registry-search/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"Equinor","countries":["NO","GB"],"maxRecordsPerCountry":5}'

from apify_client import ApifyClient

client = ApifyClient("<APIFY_TOKEN>")
run = client.actor("shelvick/global-company-registry-search").call(
    run_input={"query": "Equinor", "countries": ["NO", "GB"], "maxRecordsPerCountry": 5}
)
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    if row["outcome"] == "match":
        print(row["country"], row["legal_name"], row["registration_number"], row["lei"], row["source_url"])
    else:
        print(row["country"], row["outcome"], row["message"])

Calling from an AI agent

Agents are the primary caller: five input fields, one required, every schema description written to be constructed against without reading this page.

• Apify MCP server (mcp.apify.com) — expose shelvick/global-company-registry-search as a tool; the input schema and its per-field descriptions are advertised to the model, and the outcome + message contract lets the agent tell "this company does not exist in Finland" from "Finland could not be reached" without parsing prose. Pay per call via x402 USDC on Base or Skyfire managed tokens.
• Apify Python SDK — the snippet above, then iterate the dataset.
• REST API — POST /v2/acts/shelvick~global-company-registry-search/run-sync-get-dataset-items for a synchronous call (a full all-countries query finishes well inside the sync window), or the async /runs endpoint plus polling.

Pricing

Pay-per-event, and the event is a delivered record: one charge per matched entity record, pushed to the dataset before it is charged. Rows reporting no match, no coverage, a source failure, a rate limit or an unusable request are never charged — a query that finds a company in two countries out of eleven costs two records, not eleven lookups. A platform Actor-start event is amortized across the whole query, and maxRecordsPerCountry caps what one broad name search can bill.

See the Pricing tab on this Store page for the current per-event rates and any active subscriber discounts.

Behavior

Run-level failures. Input validation is the only run-level check, and an unusable input becomes one free invalid_input row on a SUCCEEDED run, naming the field and the fix. An unhandled fault likewise becomes one free internal_error row. The run never ends FAILED — an agent caller gets an answer, not an exception.

Per-country outcomes, all free and all carrying message: no_match (searched, nothing matched, with that register's caveat appended where it has one), not_covered (an unrecognised code, Lithuania, or a name query sent to Poland), failed (the register could not be reached or read — retryable) and rate_limited (the register's own published request limit was reached — a 429 fails that country fast and is not retried in-run, while every other country still delivers).

Latency. Countries are searched concurrently, each bounded at about 45 seconds independent of the run timeout, so one slow register cannot hold a finished query hostage. A full all-countries query typically completes in 35–40 seconds. Records are delivered as each country finishes, not in the order you listed them — sort on country if you need a fixed order.

Telemetry: to improve coverage and reliability, this Actor reports anonymous usage metrics and diagnostic events to the developer — the shape of the request (how many countries, which flags were set), the per-country outcome funnel (matched / no-match / failed / rate-limited counts), which normalized fields the delivered records actually carried, whether each match came from a number or a name, and how long the run took — and, only when something goes wrong, the query that triggered it. No account identifiers are collected, and telemetry never affects a run.

FAQ

Am I charged when a country finds nothing? No. Only outcome: "match" rows bill, and they are pushed to the dataset before they are charged. A run where every country comes back empty costs nothing beyond the Actor-start event.

Why is Poland number-only, and Lithuania not searched at all? Poland's KRS serves entity lookups by KRS number, but its name-search front end is not reachable without an interactive browser session, so a Polish name query returns a free row telling you to supply the number. Lithuania's public search moved behind an interactive browser challenge this version does not use; it stays in the coverage list with a free row explaining itself, because "we cannot search this yet" and "your country code was wrong" are different answers.

Why do most countries return an empty officers list? Because their register does not publish officer names under its open-data licence. Officers come back only from Companies House, Annuaire des Entreprises and Brønnøysundregistrene, and even there only the name, role and appointment date the register publishes. Every empty list carries an officers_note saying which case it is.

Is the data live? Every lookup hits the register at request time, so records are as current as that register publishes — except for the two documented cases above: the Singapore dataset is a monthly snapshot, and the Swiss dataset lists currently-registered entities only.

What this doesn't do

• No financial statements, filings or documents. The entity record — identity, status, form, address, activity, officers — not accounts, annual returns or filing images.
• No beneficial-ownership or shareholder data. Officers as the register publishes them, in three countries; nothing about ownership chains.
• No personal contact or identity data, ever. No dates of birth, phone numbers or email addresses, even where a source exposes them.
• No monitoring. A run answers a point-in-time question; it does not watch an entity for status changes.
• Not every country. 11 searchable today (Lithuania listed but deferred); anything outside it returns an explicit, free not_covered row rather than an empty result.

Use a filings-and-financials source for accounts and annual returns, an ownership-data provider for shareholder and beneficial-ownership chains, and a sanctions- or watchlist-screening service for the screening decision itself — this Actor produces the verified entity identity and LEI those steps take as input. For US companies use the sibling Actor below; no US register is in this coverage list.

Records come from official public registers as published, for lawful business, compliance and diligence use. Registers lag recent filings and occasionally carry clerk-side errors, so verify against source_url before a regulatory decision rests on a field.

Related Actors

Actor	Use it when
Secretary of State Business Entity Search + UCC Liens	the company is a US entity — you need state Secretary of State registration plus the UCC liens filed against it