Company Registries by Country: Every Registry Your Agent Can Query

To see the company registry by country coverage your software can read, start with the numbers: NeuralVerge's catalog lists 93 corporate registry sources that read the official company registers of 35 countries, plus two global Legal Entity Identifier directories. Each one is reachable over API or MCP and returns structured JSON. This page groups them by region, with what each register returns, the tool name and the points a lookup costs, so you can see what your agent can query before you build against it.

The short version

• —The answer. 93 sources, 35 countries, plus two global LEI directories. A source is one kind of question asked of one register: a company lookup, a name search or a list of filings. The concept itself is covered in the corporate registries guide; this page is the map.
• —What you get. The same request and response envelope everywhere, with fields that follow what each register publishes. Depth is uneven: some registers return shareholders and owners, others little beyond name, status and legal form.
• —Who it is for. Teams doing vendor onboarding and know-your-business checks, tracing ownership across borders, or monitoring counterparties in the countries they trade in.
• —What it costs. A company lookup is 5 points ($5 per 1,000) for most registers, up to 30 points for the UK and France. Name searches are 1 point per returned record in most countries. Lookups that find nothing are not billed.
• —Where the detail lives. Each source page is the reference for exact fields, request shape and prices.

What the 93 sources are

Sources are not countries

The catalog counts sources, and a source is one kind of question asked of one register. For most countries there are two sources, and for many there are three:

• —A company lookup. A registration number (or register link) returns one company as structured JSON.
• —A name search. A name returns candidate records with their identifiers.
• —A list of filings or documents. A registration number returns the register's filed documents or announcements, newest first. Not every country has one.

Across the catalog that comes to 37 company lookups, 37 name searches, 18 filing or document lists and one relationship list for LEI subsidiaries. That is the 93.

Registers differ in what they hold

Some registers publish shareholders and beneficial owners, some little beyond name, status and legal form. The request style and response envelope are uniform; the content of the result is not.

How a lookup works

Every registry source follows the same few steps, whichever country it points at.

1. Resolve the identifier

Each country has its own identifier: a Business ID in Finland, a CVR number in Denmark, a company number in the UK, an IČO in Czechia and Slovakia, a CNPJ in Brazil, an ABN in Australia. If you do not have it, the country's name search returns candidates with their identifiers, priced per returned record.

2. Send one request

The call carries the identifier in a company parameter (or a keyword and max_items for a search). The endpoint name follows the country code, such as run-fi-companies for Finland, and the parameters are in the API documentation. The same request is available as an MCP tool, with underscores in the name, such as run_fi_companies.

3. Get a typed result

The platform reads the register at that moment. The response has a stable envelope: a session id, a short readable summary, the structured result and the points charged. The fields inside the result are those of the country's register.

4. Pay only for what resolved

A company lookup that finds the record costs a fixed number of points. One that finds nothing costs nothing, and the same goes for a search with no candidates.

Company registry by country, region by region

The tables below list every registry source, grouped the way a team usually thinks about it. Each row gives the kind of register and the MCP tool and points for each question, from the source pages, counted at the time of writing. REST paths use hyphens instead of underscores. For the full field list of any country, open its page in the source catalog.

Western Europe (6 countries, 14 sources)

Austria's record is light.

The Nordics (4 countries, 11 sources)

Finland is the worked example in the next section.

Baltics, Poland and Czechia (4 countries, 12 sources)

Estonia is one of the richest records in the catalog. Czechia's list reads the Collection of Documents.

Central and South-East Europe (5 countries, 13 sources)

Bulgaria has the longest field list in the region.

Southern Europe, Cyprus and Georgia (4 countries, 10 sources)

Portugal is built from the publications portal, so it is thinner than a full extract. Georgia returns only name, status and legal form: use it to confirm existence, not to profile ownership.

The Americas (7 countries, 16 sources)

Canada covers federal corporations only. Mexico is read through published notices rather than a record per company, so its lookup takes a publication number.

Asia-Pacific (5 countries, 12 sources)

Australia's source returns business and trading names, not shareholders.

Global: LEI directories

The Legal Entity Identifier is a 20-character code that identifies a legal entity in any country. Two sources read LEI data: the Global LEI Index, and a public LEI directory that also has a relationship list for direct and ultimate subsidiaries. LEI records exist only for entities that have registered for a code, so they supplement a national register and do not replace it.

REST and MCP

Every registry source is available over REST and as a tool on the hosted MCP server, so the same lookup works from a script, a backend job or an agent in Claude. Tool names use underscores, REST paths use hyphens, and the parameters are identical. The setup for Claude and other MCP clients is in the MCP server guide. AI research, by contrast, runs over REST only.

One real request and response: Finland

To make this concrete, here is the Finnish company lookup, built from the sample on its source page. The sample is a real captured response for a listed company, Nokia Oyj, so it shows a populated record rather than a made-up one.

The request, sent to the REST endpoint with an API key:

curl -X POST https://api.neuralverge.ai/functions/v1/run-fi-companies \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"company":"0112038-9"}'

The same call as an MCP tool call:

{ "tool": "run_fi_companies", "arguments": {"company": "0112038-9"} }

The response, shortened for reading. Lines marked ... stand for fields and array entries that were removed here and are present in the full response:

{
  "session_id": "11111111-1111-1111-1111-111111111111",
  "kind": "fi_company",
  "company": "0112038-9",
  "company_id": "0112038-9",
  "human": "# Finland — PRH / YTJ — company\n\n- **Number:** 0112038-9\n- **Pages opened:** 1\n- **Proxy traffic:** 51.3 KB\n- **Name:** Nokia Oyj\n- **Status:** Active\n- **Type:** Public limited company\n- **Registered:** 1896-12-19\n- **Address:** Karakaari 7 02610 ESPOO",
  "machine": {
    "company": {
      "registration_number": "0112038-9",
      "name": "Nokia Oyj",
      "status": "Active",
      "address": "Karakaari 7 02610 ESPOO",
      "url": "https://tietopalvelu.ytj.fi/en/yritys/0112038-9",
      "company_type": "Public limited company",
      "incorporation_date": "1896-12-19",
      "activity": "Activities of head offices",
      "activity_code": "70100",
      "termination_date": null,
      "website": "www.nokia.com",
      "eu_id": "FIFPRO.0112038-9",
      "previous_names": [
        { "value": "Oy Nokia Ab", "start_date": "1966-06-10", "end_date": "1997-08-31" }
      ],
      ...
    },
    "blocked": false,
    "pages_opened": 1
  },
  "total_points": 5
}

What to notice: the result sits under machine.company, with the register link in url. The record carries previous and parallel names, which helps when matching a name from a contract. The tax debt fields are text from the register, not figures, and total_points is 5, which the source page states as $5 per 1,000. Other countries return a different machine.company object inside the same envelope.

Cost and limits

What it costs

Every price comes from the individual source pages and matches the app, and the tables above show each one.

• —Company lookups. A flat number of points per resolved record. Twenty-four of the 37 lookups cost 5 points, which is $5 per 1,000. Six cost 10, five cost 15, and the UK and France cost 30.
• —Name searches and lists. 1 point per returned record, except Portugal and Spain searches and the Portugal list, which are 2.
• —Misses. A lookup that finds nothing is not billed, and a search that returns nothing costs nothing.

Points come out of your plan's monthly allowance. See the pricing page for the plans.

Where the edges are

• —Depth is uneven. A Bulgarian or Estonian lookup returns a long list of fields. A Georgian or Austrian one returns very little. If a decision depends on ownership or financials, check the source page first.
• —Some registers are partial views. Portugal is read through a publications portal, Mexico through published notices, Spain through open data, and Canada covers federal corporations.
• —Some countries are missing. Large economies outside the 35 need another route, such as AI research or a local provider.

How to decide

Run a few companies you already know and compare the result with the register by hand. Then ask:

• —Is your country on the list, and is it the right register? Some countries have several registers for different kinds of entity.
• —Does it publish the field you need? Read the country's field list on its source page, not the headline.
• —Do you need filings or only the current record? Not every country has a filings list. Monitoring means running the same lookup on a schedule and comparing the filings list with the last run.
• —Is ownership the point? Beneficial ownership tracing works where the registers publish shareholders, as Estonia, Latvia, Bulgaria and New Zealand do.
• —Is this vendor or know-your-business onboarding? Most registers confirm a counterparty exists, is active and has the legal form it claims.
• —What does your volume cost? A 30-point lookup and a 5-point lookup are a six-fold difference at volume.

Frequently asked questions

How many registries does NeuralVerge cover?

The catalog lists 93 corporate registry sources. They read the official company registers of 35 countries, plus two global Legal Entity Identifier directories. Most registers have a company lookup and a name search, and some also have a list of filings or documents, which is why there are more sources than countries.

Do all registries return the same fields?

No. Each source returns what that register publishes. Estonia's register includes shareholders and beneficial owners, Denmark's includes employment figures, Georgia's returns name, status and legal form. The request and the response envelope are the same everywhere, but the fields inside the result follow the register.

What does a registry lookup cost?

A company lookup is a flat number of points per resolved record, from 5 points ($5 per 1,000) for most registers up to 30 points for the UK and France. Name searches cost 1 point per returned record in most countries, and 2 in Portugal and Spain. Lookups that find nothing are not billed.

Can I search by company name or do I need a registration number?

Both are available. The company lookups take the registration number the country uses, or the link to the record on the register. Every country has a separate search source that takes a name and returns candidate records with their identifiers, so you can resolve a name first and look up the record second.

Are the registry tools available over MCP?

Yes. Each registry source is available over REST and as an MCP tool, named with underscores, such as run_fi_companies for Finland.

What if my country is not on the list?

Then there is no dedicated registry source for it yet. The LEI sources cover entities that hold a Legal Entity Identifier in any country, and AI research can look for a company's footprint elsewhere, but neither replaces reading the national register itself.