Record URLs are case-sensitive and cannot be assembled by hand, so get_object_url is the only supported route and search_and_get_first_url does the search and the lookup in one step. Know one limitation going in: the Open Access API is not currently returning image URLs, so objects come back showing zero images even when you filter for image content — the rest of the metadata is intact.
An MCP server for the Smithsonian Institution's Open Access collections, giving an assistant search, object detail, exhibition status and per-museum statistics.
- Search across 3+ million objects, with filters and museum-specific queries — search_collections, search_by_unit
- Diverse sampling across museums and object types, and a way to go deeper without repeats — simple_explore, continue_explore
- Object detail and validated URLs — get_object_details, get_object_url, search_and_get_first_url
- What is currently on physical exhibit — get_objects_on_view, check_object_on_view
- What each museum actually holds — get_museum_collection_types, check_museum_has_object_type
- Museum lists and collection statistics with per-museum breakdowns — get_smithsonian_units, get_collection_statistics
A free API key from api.data.gov, set as SMITHSONIAN_API_KEY. Run it with npx -y @molanojustin/smithsonian-mcp, or install globally with npm; a Python entry point exists as well. Version 1.2.9.
One command plus a key — npm install -g @molanojustin/smithsonian-mcp, then supply credentials
