Integrating an application

A checklist for using the register in an invoicing or accounting application.

Let your AI agent integrate company search and lookup into your application, following these rules.

The pattern that works for an application such as an invoicing tool:

Search proposes

Show /v1/search results as the user types, with a debounce. Show the name, the IČO and the full address: two subjects can share a name.

Only an exact identifier resolves

When the user picks a result, call /v1/lookup/{ico} and use that answer. Never fill a form from the first search result on your own.

Keep the versions

Store each section’s source.version with your copy of the counterparty.

Check daily

Once a day, call /v1/lookup:batch with fields: "versions" and show the user what changed. See Change detection.

Rules that keep documents correct

Unknown is not “not found”. Treat 503, unknown and stale as “unknown”. An outage of the register must not block issuing a document; let the user continue with the data they confirm.
VAT wording. Print “platiteľ DPH od …” only when vat.payer is true. §7 and §7a subjects have an IČ DPH but are not VAT payers. See VAT status.
The register entry pre-fills; the user confirms. identity.registrations[] gives the court and number for a supplier’s legal line. The genitive court name and the exact wording are yours to build.
Cite the date. “According to the Financial Administration list of 10 Nov 2026” is accurate; “is a VAT payer” without a date is not.
EU VAT numbers. Query VIES yourself: the register does not mirror it.
Attribution. If you show the data to others, show the attribution texts the licences require.

Retry safe GET requests after a connection error, with a backoff.
Honour Retry-After on 429.
Read /v1/capabilities at start-up: whether fuzzy search is enabled, and the limits.