Smithsonian Open Access Scraper

Search and browse the Smithsonian Institution's Open Access collection — over 2.8 million CC0 public-domain images and object records spanning 30+ Smithsonian museums, archives, and research units (National Portrait Gallery, National Air and Space Museum, Smithsonian American Art Museum, National Museum of Natural History departments, Cooper Hewitt, Hirshhorn, and more). Every record returned is public domain (CC0) — free to use, remix, and republish with no permission needed.

What this actor does

• Three modes: search (keyword), byUnit (browse a specific museum/unit), byObjectType (browse by kind of object)
• 30+ museums and units covered, from fine art to natural history specimens
• Real CC0 images: thumbnail and high-resolution download URLs for every item that has one
• Rich metadata: maker/artist, date, medium, dimensions, credit line, topics, object number
• Empty fields are omitted — every record only contains what the Smithsonian actually published for that item

Output per item

• id, recordUrl, guid — identifiers and canonical Smithsonian page link
• title
• unitCode, unitName — owning museum/unit
• makers[], primaryMaker — artist / maker / sitter names
• date, datePeriods[] — display date and decade/period buckets
• objectType, objectTypes[] — kind of object (e.g. Photograph, Painting, Specimen)
• topics[] — subject/topic tags
• medium, dimensions[]
• creditLine, rights (almost always CC0), objectNumber, collection
• dataSource, notes (collection/curatorial description), place (geographic origin), publisher (related publication, when applicable)
• imageUrls[], thumbnailUrl, highResImageUrl, mediaCount
• recordType: "object", scrapedAt

Input

Field	Type	Default	Description
mode	string	search	search / byUnit / byObjectType
searchQuery	string	portrait	Keyword matched against title, maker names, and topics
unitCode	string (select)	npg	Museum/unit to browse or scope to (mode=byUnit required)
objectType	string	Photograph	Object type match, e.g. Painting, Sculpture, Coin, Specimen (mode=byObjectType required)
onlyWithImages	boolean	true	Only emit items that have a downloadable image
maxItems	int	25	Hard cap (1–500)

Example: search across all units

{ "mode": "search", "searchQuery": "moon landing", "maxItems": 20 }

Example: browse the National Air and Space Museum

{ "mode": "byUnit", "unitCode": "nasm", "maxItems": 50 }

Example: find paintings anywhere in the collection

{ "mode": "byObjectType", "objectType": "painting", "maxItems": 30 }

Use cases

• Digital archives & research — bulk-collect CC0 images and metadata for a topic or museum
• Education — source public-domain historical images for course materials
• Creative reuse — royalty-free imagery for design, publishing, or generative-AI training sets
• Museum data analysis — compare object types, makers, and eras across Smithsonian units
• Content discovery — find high-resolution scans of specific artifacts, artworks, or specimens

Data source / limitations

This actor reads the Smithsonian's public Open Access bulk metadata (announced via the Smithsonian/OpenAccess GitHub repository), which is mirrored as line-delimited JSON files on a public AWS S3 bucket, organized by owning unit and content hash. There is no interactive search index over the full 2.8M+ records without downloading the entire ~47 GB dataset, so:

• search and byObjectType scan a bounded, round-robined sample of shard files across the relevant unit(s) rather than the full corpus. Broad or common terms (e.g. "portrait", "airplane") reliably return results; very obscure terms may return fewer matches than a true full-text index would find. Scoping to a specific unitCode speeds this up and improves relevance.
• byUnit walks a unit's shard files sequentially and is the most exhaustive, reliable mode.
• Archive-focused units (Archives of American Art, National Anthropological Archives, Human Studies Film Archives, Eliot Elisofon Photographic Archives, Freer and Sackler Archives, Anacostia Community Museum Archives) primarily hold archival finding-aid metadata rather than individually digitized media records, so they may return fewer image-bearing items than art or history museums.
• recordUrl links to the item's page on its owning museum's own website (e.g. npg.si.edu, airandspace.si.edu). Image URLs (imageUrls, thumbnailUrl, highResImageUrl) are served from the Smithsonian's shared ids.si.edu media CDN and are always publicly reachable. A handful of individual museum websites (observed: National Portrait Gallery's npg.si.edu) run their own bot-protection in front of the human-facing object page, which can reject fully automated/scripted requests even though the same link opens normally in a real browser — this only affects clicking through to that one museum's page, never the metadata or images themselves.
• No login, API key, or paid proxy is required — the bucket and all image CDN links are fully public.

FAQ

Is this data really public domain? Yes — the Smithsonian released this collection under CC0 in 2020. No attribution is legally required, though crediting the Smithsonian is appreciated.

Why do some items have no image? Not every cataloged object has been digitized yet, and some records (especially in archival units) describe a folder or document rather than a single photographed object. Set onlyWithImages: true (the default) to skip these.

What's the difference between objectType and objectTypes? objectType is the primary/first type Smithsonian assigned to the item; objectTypes is the full list when an item has more than one classification.

Can I get every record for a museum? Run mode=byUnit with a high maxItems. Very large units may require multiple runs to page through fully, since each run samples a bounded number of shard files.

Are the image URLs safe to hotlink? Yes — they're served from the Smithsonian's own ids.si.edu media delivery service and are publicly accessible without cookies or a referer header.