← Guides
Guide

Build your own prospect list from open company registers, with Claude

I had Claude pull the open company registers of country after country into one database, and it now holds over 53 million active companies. You don't need 53 million rows to find your prospects. For your markets and one kind of customer, a couple of days of prompting gets you a really good list to start from, and this guide is how I would do it.

It is written for someone who sells to companies (an adviser, a broker, a consultancy, a sales lead at a small firm) and already pays for Claude or ChatGPT. You need no software of ours.

Why start from the register

Most of what prospect databases sell is public data: financial documents that are publicly available, financial numbers, news they scrape, board members and leadership positions, sometimes with some manual cleanup on top. I've evaluated a couple of them in earlier jobs, and they would say their data is better, and for some of it that is probably true. But you need data that is specific to your industry and your region, and there is so little chance that a general database fits your particular list of prospects that you'd better research them yourself anyway. So get the register itself, the high level, and from there Claude can do the research on its own, which is what you want anyway.

What it looks like across the Nordics

To show the steps I ran them on 15 November 2026 for a made-up seller who sells to management consultancies in Sweden, Norway and Finland. All three registers are open as one download each.

Sweden Norway Finland
The whole register, one file 1,829,348 1,181,339 466,345
Main industry code 70.20, management consultancy 82,720 32,171 24,793
Limited companies (AB, AS, Oy) 65,840 17,568 24,131
Active, by the register's own rule 62,602 17,191 23,655
With an employee count in the register none 5,633 none
Website in the register none some 4,417

That is 103,448 active consultancies in three countries. The three downloads took about a minute side by side and the filtering seconds.

The table also shows what changes when you cross a border. Each file has its own format. "Active" means three different things: in Sweden that the company is registered for VAT, tax or as an employer, in Norway that it isn't bankrupt or being wound up, in Finland that it is registered with no bankruptcy or liquidation on file. Only Norway's file says how many people work there, so the size filter you'd use in one country is gone in the next. And Denmark's register is public, but I found no open whole-register file to download the same way.

The steps

None of this needs a clever prompt. Each step is just asking Claude, in plain words, to find the public company data for your markets, download it and relate it to who you sell to, and it will just do it. These are the asks I'd start from.

1. Say who you sell to in the register's terms. A register knows a company's industry code, legal form, address, founding date and sometimes its size. It doesn't know whether they need you.

I sell [what] to [who] in [countries]. Which fields in their company
registers (industry codes, legal form, region, size) describe them?
And what about them will no register tell me?

Keep the second answer, because it is what the research in step 5 is for.

2. Find the register for each market. Ask for the official source, because the first page of results for any register is full of companies selling its data back to you:

For each of [countries], find the official company register, whether
it has a free download or an open API, and what its licence says.
Link the register's own page.

Some that are open as files: Sweden's company file from SCB, published through Bolagsverket, Norway's Enhetsregisteret (the whole register as one CSV or JSON file), Finland's PRH open data (updated daily, free), the UK's Companies House Free Company Data Product (a monthly snapshot of live companies as CSV), and for listed US companies SEC EDGAR. Terms differ between countries and some charge for bulk files or need an agreement, so read the register's own page before you rely on it.

3. Download them and filter them. A whole-country file is too big to paste into a chat, so use Claude Code, or any Claude that can run code on your computer:

Download those registers and filter each to the companies from step 1.
Show me the industry codes and the status values in each file first,
then the count after every filter, and save one spreadsheet with a
column for the country.

Asking for the codes and statuses first saves you a wrong turn. In Norway's file the consultancy code is 70.200, and not one company carries 70.220, the code I would have guessed. In Sweden's file a company's status is a number, and you need the statistics office's own description to know that 1 means active.

4. Check what the registers get wrong before you trust them. Size is often missing: Sweden's and Finland's open files have no employee count at all, and in Norway only 5,633 of 17,191 active consultancies have one, so a size filter quietly drops most of the market. Revenue, where a register has it, is usually the parent company's own figure, not the group's. And matching companies by name finds the wrong one: when we screened the listed companies in Stockholm, matching by name picked the wrong company for 28 of 233, and the register had no revenue at all for 61 (the whole screen is here). Join everything by registration number and country, never by name.

90% off Rather skip steps 1 to 4? Start from 53 million companies. ↓

5. Let Claude research the shortlist. This is where the list becomes yours, and where step 1's second answer comes back:

For each company in this spreadsheet, find its website, what it does,
its latest annual report and this year's news, and score it 1 to 5 on
[what no register holds], with the source beside every score.

With 103,448 rows you narrow first, by city or by what step 1 said, then start with 30 or so, read what comes back, and fix the ask before you run the rest.

What it won't give you

A register gives you companies, not people. Who to call at each one is still your research (and names and emails of people are personal data, with rules of their own). Its size and revenue fields are patchy or missing, as step 4 shows, and it says nothing about whether a company needs what you sell, which is why step 5 exists. And in some countries the bulk file costs money, needs an agreement or isn't there as one file at all.

What I did with the company database

I kept improving it, and I never stopped. It is now the data layer in The Revenue Assistant: over 53 million active companies, with 32 countries from their own official registers (the Nordics among them, Denmark included) and the companies with an LEI in 193 more. It is one format for every country, with company size for over 6 million, and you reach it through an API or an MCP server from your own AI tools. Because of the sheer size of it, it costs me quite a lot to set up and host, but I'll keep improving it over time.

If you'd rather not build your own, or your market spans several countries, hook up to it as your first query and let Claude research the shortlist from there. The Revenue Assistant is an app for Mac and Windows that runs on your team's own Claude or ChatGPT, and you can buy the data without the app.

Something here unclear or wrong? Write to us — it reaches the person who wrote it.
Augmented Mind AB · PB 13, 00561 Helsinki, Finland