Art Institute Chicago MCP Server

What art-institute-chicago-mcp-server MCP server does

The art-institute-chicago-mcp-server MCP server exposes the Art Institute of Chicago’s public API to MCP-compatible clients. It covers a collection of about 133,000 artworks, along with artist records, exhibitions, and audio-guide stops.

Artwork search accepts text and filters such as artist, department, material, technique, subject, gallery, date range, public-domain status, on-view status, and image availability. Results can include facet counts and can be ordered by relevance or date. A separate retrieval tool returns fuller records, including descriptions, provenance, exhibition and publication history, catalogue information, image URLs with rights status, and related media.

The server also supports artist and organization lookup, exhibition searches across past, current, and upcoming shows, transcript and title searches for audio-guide stops, and vocabulary lookup for valid artwork filter values. An artic://artworks/{id} resource provides an individual artwork record as JSON.

How it works

Requests are sent through MCP to the museum’s public API. The server can run as a stdio process or as a Streamable HTTP service. A hosted endpoint is available at https://art-institute-chicago.caseyjhand.com/mcp, while local installations can use the published npm package.

Searches use bounded pages and report the applied limits. Artwork results are paged within the first 1,000 matches, with up to 12 rows per page. Artist and exhibition searches allow up to 25 results per page, audio-guide searches return up to two stops per page, and vocabulary lookups return up to 100 values. Larger retrieval requests may report deferred IDs for a later call.

Setup and configuration

The art-institute-chicago-mcp-server MCP server does not require an API key according to the project documentation. A local stdio configuration can launch the npm package with npx -y @cyanheads/art-institute-chicago-mcp-server. The repository also provides a Claude Desktop package, installation links for Cursor and VS Code, and a Docker container reference.

Choose the hosted endpoint when the client supports Streamable HTTP and you do not need a local process. Use the stdio package for desktop or editor configurations that launch MCP commands directly. No environment variables are documented as required.

Tools and capabilities

• artic_search_artworks: Search by text and collection filters, sort by date, and request facets.
• artic_get_artworks: Retrieve up to 10 attempted artwork records per call, with selectable sections and related media.
• artic_search_artists: Find artists, cultures, and organizations by name or museum ID, including life dates, artwork counts, sample works, and available Getty ULAN identifiers.
• artic_search_exhibitions: Filter exhibitions by text, date range, timing, artwork ID, or artist ID.
• artic_search_audio_guide: Find guide stops by title or transcript and return MP3 URLs and transcript text.
• artic_lookup_vocabulary: Enumerate accepted filter values with artwork counts.

Limitations and notes

The museum’s artist relationship data is not exhaustive, so a missing recorded link does not prove that an artist or artwork never appeared in an exhibition. Getty ULAN coverage is also sparse and uses identifiers distinct from the museum’s own IDs.

Audio-guide responses include licensing and source information and are designated for noncommercial educational and personal use. Artwork descriptions may include attribution requirements, and image rights status should be checked before reuse. Unknown artwork resource IDs fail with an artwork-not-found error.