Europeana
Europeana is the single front door to Europe's digitised cultural heritage. It aggregates descriptions of paintings, books, maps, manuscripts, photographs, sound recordings, film, and 3D objects from thousands of museums, libraries, galleries, and archives across the continent, and presents them through one multilingual catalogue. The scale is in the tens of millions of items drawn from around 4,000 institutions, so it is less a single collection than an index over many.
The key thing to understand is the split between metadata and content. The metadata (titles, creators, dates, places, subjects, descriptions, and the links between records) is the layer you can use immediately, and it is released under CC0, meaning it is effectively public domain with no conditions. The actual digital objects (the image files, the scanned pages, and the audio) carry their own rights statements that vary item by item, from fully open to all-rights-reserved. Europeana is honest about this and labels each item, but it means you cannot assume the picture is as free as the record describing it.
There are several access routes. The Search API returns metadata records and media references, the Record API gives full detail on a single item, and there is IIIF support (a shared standard for serving high-resolution images so viewers can zoom and compare across institutions). For bulk work, use the OAI-PMH harvesting endpoints to pull metadata at scale, and the SPARQL endpoint if you want to query it as linked data and follow connections out to VIAF, GeoNames, Wikidata, and DBpedia.
For RAG, the metadata is what you index: it is clean, structured, multilingual, and safe to reuse, which suits cultural-heritage search assistants, education tools, and provenance or exhibition research. Treat it as a discovery and description layer rather than a full-text corpus, since most records do not include the transcribed text of the object itself. If you need American collections alongside it, the Digital Public Library of America follows a very similar metadata-first model.
Related sources
Common European Data Space for Cultural Heritage
The umbrella project coordinating cultural heritage data across Europe. Led by the Europeana Initiative, it helps cultural institutions and EU Member States share their collections through a common, shared infrastructure.
Digital Public Library of America (DPLA)
An aggregator that brings together content from America's museums, libraries, and archives. Its API offers metadata on individual items and on collections, and the whole repository is available to download as zipped JSON files.
HathiTrust Digital Library
A partnership of academic and research institutions offering millions of digitised titles. Its Research Center provides computational access to the full corpus, including in-copyright works, under non-consumptive research terms.
Smithsonian Open Access
Millions of digital items from the Smithsonian's museums, research centres, and archives, released under CC0. Includes images, 3D models, and research datasets you can reuse freely, including commercially.