Questions

No. A search engine finds pages. We build the full record of a company, every company linked to it and the people behind them, across thirty years of public sources, connected and verified. You query it from your own AI, or work with our analysts.

Four kinds of links. Ownership: shareholders, subsidiaries, holdings, ultimate owners. People: directors, founders, officers, and their other companies. Business: suppliers, clients, partners, lenders, public contracts. Shared traces: common addresses, domains, servers, contact details.

We follow these links several degrees out from the company, and do the same for every company we reach.

Depth. AI search reads a few surface pages. It misses registry filings, hours of recordings and deleted pages, confuses people who share a name, and often repeats a company’s own story. The record has already read and structured these sources, and every answer points to the document behind it.

These platforms organise official company data: registries, filings, accounts, officers, often country by country. We start where they stop. We add what was said in interviews, podcasts and video, what was removed from the web, courts, patents, public money, property and more, and we link all of it across companies and people.

Twenty layers of the public record, from the open web, news and company filings to courts, sanctions, macroeconomic series and laws. Plus licensed data from established providers.

Nothing is hacked, nothing comes from leaked datasets, and no private messages are collected.

The twenty sources →

We work across jurisdictions. What is made public varies from one country to another, and the depth of the record varies with it.

Each source goes as far back as it exists, some about thirty years. Records are kept even after they disappear from the live web. Sources refresh at different rates, and every answer shows the dates of the records it relies on.

Yes. Every statement in the record points to its original document, with its date. When two sources contradict each other, both are kept and the contradiction is flagged.

Public audio and video are transcribed with language models and our own algorithms. Speakers are distinguished, and each statement is attributed to its speaker, its source and its time. Transcripts are linked to the rest of the record, so what someone said can be compared with what was filed.

The record is built first. Then any question can be asked of it, as deep as the network goes. You can work with our analysts, query it from your own AI, or test your assumptions and simulations: change a condition and see where it travels.

Any AI that supports MCP, the open standard used by most major assistants and agent tools. You add it as a connector with the access details we provide: it takes a few minutes and mostly means pasting a link. Developers can plug it into their own agents. We help with the setup.

In part. Agents check each link against its sources, and our analysts review a share of the findings.

Yes. The connection can be shared across your team, or we deliver the record for you to run in your own environment. We agree the scope with you.

Yes. We sign your NDA or ours, and complete your vendor questionnaire. We never access your systems and only receive lookups, so onboarding is usually light. For the strictest environments, we deliver the record as encrypted files (PGP, with your own key), over SFTP or to your cloud storage, to run on your own infrastructure.

Sources are read in their original language. We work and deliver in English and French.

No, we don’t see them. Your questions stay in your own AI. The record only receives lookups, such as a company or a person’s name, over an encrypted connection. Lookups are kept 30 days for security, then deleted. They are never shared, sold or combined across clients.

We sign an NDA before you share anything sensitive. For sensitive work, we deliver the record as encrypted files that only your key can open, to run entirely on your own infrastructure.

We publish our own research. We never publish anything about our clients’ work, and we never disclose who our clients are.

It is not a KYC or background-check tool. It is built to understand companies and the people who run them, in their public and professional roles.

It may not be used to harass, track or locate private individuals; to score or profile people; to screen job candidates, tenants, borrowers or insurance applicants; to discriminate on protected characteristics; or to break the law. A breach ends access.

It is designed to. It holds public information about people in their public and professional roles, processed under legitimate interest (Article 6(1)(f) GDPR) and within the limits above.

Anyone can ask what is held about them, request a correction or removal, or object to processing, through the Data requests page. We reply within one month. Company-level data, such as registry records, filings or court decisions, comes from official public records; if something is inaccurate, write to us and we will review it.