EU Open Data Catalogue Directory

Who publishes the EU open data, and how much of it. One row per catalogue that data.europa.eu harvests: the id, the title, the country, the homepage, and how many datasets the portal holds for it. Measured on 2026-09-17: 211 catalogues.

Two reasons to have this as rows instead of a web page:

1. It is the key to the other Actors. Every EU open data Actor in this family takes a catalogue filter, and the ids are not guessable — data-gov-be, govdata, datos-gob-es. This is the list you pick them from.
2. It is a coverage map. Which national portals does the EU actually harvest, which of them are empty, and which one is behind the 29,602 datasets you keep hitting? That is a join the portal does not publish as a table.

What a row looks like

{
  "id": "data-gov-be",
  "title": "data.gov.be",
  "description": "Belgian open data portal, harvesting data from federal, regional and local portals in Belgium",
  "portal_url": "https://data.europa.eu/data/catalogues/data-gov-be?locale=en",
  "homepage": "https://data.gov.be",
  "country_code": "BE",
  "country_label": "Belgium",
  "dataset_count": 19000,
  "source_type": "dcat-ap",
  "languages": ["fr"],
  "issued": "2024-08-30T08:43:47Z",
  "modified": "2024-08-30T08:43:47Z",
  "publisher": null,
  "publisher_withheld": true,
  "child_catalogue_ids": [],
  "child_catalogue_count": 0
}

Live API, 2026-09-17.

Input

Every field is optional. With no input you get the whole directory, biggest catalogue first.

Field	What it does
query	full-text filter over catalogue title and description
country	lowercase ISO codes, as the portal spells them: be, de, fr
source_type	how the catalogue is harvested: dcat-ap, euodp-legacy, …
min_datasets	drop the catalogues the portal holds nothing for
sort_by	datasets (default, largest first), title, or modified
max_results	1 to 500
language	which language to collapse multilingual titles to. Default en

min_datasets and sort_by are applied to the records after they are fetched, not by the portal. That is why total_matches (what the portal matched) and returned (what survived the local filter) can differ, and why both are in the output.

publisher_withheld, and why a field is sometimes empty on purpose

The portal types a catalogue's publisher either as a corporate-body classification — which proves it is an organisation — or as the bare DCAT class Agent, which also covers people. Measured over 100 catalogues on 2026-09-17: 61 proven, 39 unproven.

This Actor collects no personal data. When the type does not prove the publisher is an organisation, the name is withheld and publisher_withheld is true, so you can tell that apart from a catalogue that simply has no publisher (publisher_withheld: false). The title and homepage identify the portal in either case.

Licensing and attribution

Portal metadata is dedicated to the public domain under CC0 1.0 (legal notice). The source block on every run carries the terms URL and the notice, so a stored result explains its own licence.

Billing

You are charged per catalogue row delivered to your dataset. A run that matches nothing delivers no rows and costs nothing.