rijksmuseum-mcp+
This server lets you search and explore Rijksmuseum collection metadata, examine artworks and their images in detail, and retrieve scholarly, conservation, curated-set, and change-tracking information.
Search artworks by structured filters — object type, creator, title, material, technique, date, subject/person, object number, provenance availability, image availability — with sorting, facet counts, pagination, compact counting, and parent/child grouping.
Retrieve full artwork metadata for a single object: titles, creator, dates, dimensions, materials, production details, provenance text, inscriptions, subjects, themes, exhibitions, related objects, physical relations, location, licenses, and external authority IDs.
Access scholarly bibliography: get citations for an artwork (with publication, pages, ISBN, WorldCat/library links) or reverse-lookup which artworks cite a given publication.
Review conservation history: technical examinations (X-ray, infrared, dendrochronology, paint samples) and restoration/conservation treatment events.
View artworks interactively: open a deep-zoom viewer, remount it on related objects, and zoom/pan to specific regions.
Inspect images for analysis: retrieve full or cropped image bytes (with rotation, grayscale, and pixel/percentage regions) for the LLM to examine details, inscriptions, or specific areas.
Browse curated sets from Rijksmuseum staff, with set statistics and paginated member listings.
Track collection changes via an OAI-PMH delta feed of recently created, modified, or deleted records.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@rijksmuseum-mcp+ find paintings by Vermeer"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
rijksmuseum-mcp+
Overview
The rijksmuseum-mcp+ MCP server lets you explore the Rijksmuseum's artwork collections through natural conversation with an AI assistant. It does this by creating a bridge between the AI system's chat environment and an enriched copy of the museum's open-access, curated metadata. This in turn enables many features beyond those offered by the Rijksmuseum through their Search API and collections portal. Rijksmuseum-mcp+ works best when used together with rijksmuseum-iconclass-mcp, a sibling resource for searching and exploring Iconclass concepts.
Please do not treat the data made available by this resource as current or authoritative. It is based on data copied from the Rijksmuseum on May 2nd, 2026. For current data, please always use the Rijksmuseum's own search portal and APIs. Nor have the (in small part, LLM based) enrichments of the museum's provenance data been reviewed or endorsed by the Rijksmuseum. This is a technology demo that remains – by its nature – in active development.
This MCP server was developed by the Research and Infrastructure Support (RISE) group at the University of Basel. We are particularly interested in exploring the research opportunities, methodological risks, and technical challenges posed by retrieving and analysing data with LLMs. If you are interested in collaborating with us in this area, please get in touch.
Related MCP server: Cultural Heritage MCP Server
Key Features
Searching artwork metadata. You can search artworks by keyword, by structured filters (artist, type, material, technique, date, physical dimensions, production place), or by meaning — a semantic search that handles interpretive queries like "melancholy winter scenes at dusk". There's also iconographic search via Iconclass codes, so you can ask for works depicting a specific scene or motif rather than just matching words in titles.
✨ Searching digitised artworks visually. A multimodal search can match a plain-language description such as "children playing in the street" against the digitised images themselves, not only against catalogue metadata. Each result then notes whether it was found through the image, the catalogue text or the Iconclass codes. The search can be further narrowed with filters, for example to paintings, a period, or a single artist.
Looking closely at works. Any digitised artwork can be opened in an interactive, inline viewer. You can instruct the AI-assistant to inspect image regions by itself — useful for analysing a specific area of the image you've highlighted for it in the viewer. If an artwork has curator defined related artworks (e.g. preparatory sketches, or different impressions of the same design) these can be accessed through the viewer as well — its < / > buttons step through them in place, without leaving the viewer.
Collection-level analysis. You can ask for statistical breakdowns across the whole collection: top creators, distributions by decade, type, or theme, geographic spread, as well as demographic questions.
Provenance and ownership history. Trace who owned a work and when, which works passed through a particular collector or dealer, sales and confiscations in a given city or period, price histories, and how long families held their collections. This is made possible by an experimental AAM parser that permits structured, CMOA/PLOD-aligned queries.
Scholarly apparatus. Lookup bibliographies for individual works, reverse lookups (which artworks cite a given publication), and conservation histories including technical examinations such as X-rays, infrared, and dendrochronology.
Relationships between works. A similarity engine ("find images similar to..") compares works across multiple dimensions — visual, thematic, lineage, shared subject — and surfaces pendants, pairs, copies, reproductive prints after paintings, and different impressions of one design.
People and places. You can search persons by profession, lifespan, or birthplace and then view their works, or run geospatial queries (e.g. "works depicting places within 20 km of Haarlem").
Linked Open Data. Works carry persistent handle.net URIs and other external IDs, and entities (creators, materials, depicted persons and places, themes) carry identifiers linking them to Wikidata, VIAF, ULAN, and RKD.
Command-line interface. The bundled
rijks-mcptool runs the same queries from the terminal — each tool exposed as a verb, with JSONL output for piping into tools such asjqso that results are scriptable and reproducible.
Sample Queries
The system is designed to let you search, explore and ask questions about the Rijksmuseum's collections in natural language. For example:
What is the actual distribution of Rembrandt's works in the Rijksmuseum across painting, printmaking, and drawing — and how does this challenge popular perceptions of him as primarily a painter? Demo
For examples of more complex queries and sample responses, please browse the research scenarios. Nine scenarios for art historians and curators cover iconographic, semantic and multimodal search, image inspection, similarity search, provenance, collectors' marks, bibliography and conservation records, each with a shared demo transcript.
Quick Start
The best way to get started is with Claude Desktop or claude.ai by adding rijksmuseum-mcp+ to Claude as a remotely hosted, custom 'Connector' using the URL below. This is currently free for one connector – additional connectors require a paid ('Pro') or higher subscription from Anthropic.
https://rijksmuseum-mcp-plus-production.up.railway.app/mcpGo to Customize → Connectors → Add custom connector → Name it as you like and paste the URL into the Remote MCP Server URL field. You can ignore the Authentication section. Once the connector is configured, optionally set the permissions for its tools (e.g. 'Always allow'). See Anthropic's instructions for more detailed instructions.
Many other desktop and web-based clients such as OpenAI's ChatGPT or Mistral's Vibe and open-source applications support remotely hosted, custom MCP servers. Please consult their documentation for more information. In particular, the free version of Mistral's Vibe already has excellent support for MCP servers and skills (see below). In contrast, ChatGPT can currently (October, 2026) only support the full functionality of this server (in particular, viewing and interacting with images in the chat timeline) in conjunction with a paid Business or Enterprise subscription. So at the moment the best way to use the hosted MCP server for free is either with Claude or with Mistral. Alternatively, you can install rijksmuseum-mcp+ locally on your own computer. Please consult the technical guide for more details.
Afterwards, follow the same procedure to install rijksmuseum-mcp+'s sibling resource for IconClass, rijksmuseum-iconclass-mcp. This allows you to automatically search and explore c. 1.3 million IconClass notations, concepts, and descriptive texts alongside the Rijksmuseum's metadata.
Research skill
The rijksmuseum-mcp+ skill file (rijksmuseum-mcp-plus.skill) gives the AI assistant detailed guidance in natural language on how to use rijksmuseum-mcp+ effectively: which tool to choose for a given question type, how to combine searches, important metadata distinctions and known limitations. The package also includes reference files with full description of the available provenance search patterns and the find_similar functionality. Skills were originally developed by Anthropic for their Claude products but have since become an open standard. Making use of this skill is optional but will significantly improve the quality and efficiency of your AI assistant's responses when exploring the collection. The downloaded skill file can be installed in Claude by following these instructions.
Currently (October, 2026) ChatGPT's free tier makes uploading skill files cumbersome. Instead, use Mistral Vibe and install the skill via the 'folder' option, first downloading and unpacking rijksmuseum-mcp-plus.zip.
Technical Design
For local setup (stdio or HTTP, also via cli), deployment, architecture, data sources, and configuration, please see the technical guide. For measurements of what the server costs and saves in tokens compared with plain web search, see the (still early and experimental) benchmarks.
The enriched metadata is available as a standalone SQLite database (vocabulary.db.gz, from the v0.90 release). Its tables, joins and example queries are described in the database schema.
Roadmap
Ongoing:
fix bugs and fine-tune queries and tool descriptions
Later:
paper/presentation
Maybe:
OCR/HTR of appropriate images in the collection
incorporate historical exhibition data
integrate with other Linked Open Data resources (e.g. Colonial Collections)
support inferred geolocation data
improve the
descriptionsignal for find_similar (e.g. via a LLM re-ranker)
Authors
Arno Bosse — RISE, University of Basel with Claude Code, Anthropic.
Citation
If you use rijksmuseum-mcp+ in your research, please cite it as follows:
APA (7th ed.)
Bosse, A. (2026). rijksmuseum-mcp+ (Version 0.96) [Software]. Research and Infrastructure Support (RISE), University of Basel. https://github.com/kintopp/rijksmuseum-mcp-plus
BibTeX
@software{bosse_2026_rijksmuseum_mcp_plus,
author = {Bosse, Arno},
title = {{rijksmuseum-mcp+}},
year = {2026},
version = {0.96},
publisher = {Research and Infrastructure Support (RISE), University of Basel},
url = {https://github.com/kintopp/rijksmuseum-mcp-plus},
orcid = {0000-0003-3681-1289},
note = {Developed with Claude Code (Anthropic, \url{https://www.anthropic.com})}
}Image and Data Credits
Collection data and images are provided by the Rijksmuseum, Amsterdam via their Linked Open Data APIs.
Licensing: Information and data that are no longer (or never were) protected by copyright carry the Public Domain Mark and/or CC0 1.0. Where the Rijksmuseum holds copyright, it generally waives its rights under CC0 1.0; in cases where it does exercise copyright, materials are made available under CC BY 4.0. Materials under third-party copyright without express permission are not made available as open data. Individual licence designations appear on the collection website.
Attribution: The Rijksmuseum considers it good practice to provide attribution and/or source citation via a credit line and data citation, regardless of the licence applied. Please see the Rijksmuseum's information and data policy for the full terms.
Enriched data: The enrichments this project adds to the vocabulary database, such as the parsed provenance tables, and the embeddings database (gemma.db) are released under CC0 1.0. The Rijksmuseum records they are built on keep the licences described above.
This project was inspired by @r-huijts/rijksmuseum-mcp, the original Rijksmuseum MCP server based on the museum's now superseded REST API.
License
The software in this project is licensed under the MIT License.
Available Tools
13 toolsbrowse_setBrowse SetARead-onlyIdempotent
Enumerate the member artworks of one curated set by setSpec. Get setSpec values from list_curated_sets. DB-backed (warm calls in tens of ms). Returns DB-direct records with objectNumber, title, creator, date (display + earliest/latest), description, dimensions, datestamp, image/IIIF URLs, and a stable lodUri. For multi-row vocab (subjects, materials, type taxonomy, full set memberships), follow up with get_artwork_details on the returned objectNumber. Supports pagination via resumptionToken (stateless base64; not portable across server upgrades). Not for set discovery — use list_curated_sets first.
Inputs: setSpec, maxResults, resumptionToken, includeExtentText. Returns: error, records, totalInSet, resumptionToken
find_artworks_citing_publicationFind Artworks Citing a PublicationARead-onlyIdempotent
Reverse bibliography: artworks citing a given publication URI or id. Use the publicationUri from get_artwork_bibliography (e.g. 'https://id.rijksmuseum.nl/301154354') or the bare id. Local and resolver-free. Not for topic search of the library catalogue.
Inputs: publication (required), full. Returns: error, total, artworks, warnings, +2 more
get_artwork_bibliographyGet Artwork BibliographyARead-onlyIdempotent
Scholarly references (citations) for ONE artwork by objectNumber. Includes linked publication, pages, ISBN where known. Entries with a linked publication carry a [pub NNN] handle — pass it to find_artworks_citing_publication. Follows a search_artwork / get_artwork_details result. By default returns the first 5 plus a total count; set full=true for all entries (major works can have 100+ — mind the context window). Not for general metadata — use get_artwork_details. Not for library-catalogue search.
Inputs: objectNumber (required), full. Returns: error, total, entries, warnings, +1 more
get_artwork_detailsGet Artwork DetailsARead-onlyIdempotent
Full metadata for ONE artwork by objectNumber. Covers creator, dates, materials, provenance, inscriptions, related objects. Typically follows a search_artwork / semantic_search / find_similar result, or a user-named objectNumber. Provide exactly one of objectNumber (e.g. 'SK-C-5' for The Night Watch) or uri (a Linked Art URI from relatedObjects).
Returns metadata including titles (primary plus the full set of variants with language and qualifier — Dutch/English brief/full/display/former), creator, date, dateDisplay (free-text form), description, curatorial narrative, dimensions (text + structured: height/width/depth/weight/diameter where present), extentText, materials, object type, production details (with creator life dates, gender, and Wikidata ID where available), raw provenance text (structured chain: search_provenance), credit line, inscriptions, license, related objects (each carrying objectNumber + iiifId for in-viewer navigation), themes, exhibitions, attributionMarks (signature/inscription counts), externalIds (handle + other), location (museum room when on display, as { roomId, floor, roomName }), recordCreated/recordModified timestamps, plus collection sets and reference metadata. Authority IDs appear at two levels: work-level under externalIds (handle + other), and entity-level under equivalents[] arrays on objectTypes, materials, production entries, subjects.depictedPersons / subjects.depictedPlaces, collectionSetLabels and themes — each entry a { authority, id, uri } triple (VIAF/ULAN/RKD/Getty TGN+AAT/GeoNames/Wikidata), and one entity can carry several, so read every entry (iconclass terms have none). The relatedObjects field carries each peer's objectNumber (canonical handle) plus a Linked Art objectUri; pass either form back here, objectNumber preferred.
Not for filter-based discovery — use search_artwork. Not for similarity discovery — use find_similar. Not for aggregate counts — use collection_stats.
Inputs: uri, objectNumber, verboseExtent. Returns: id, url, date, type, +44 more
get_artwork_imageGet Artwork ImageA
Opens an interactive deep-zoom viewer for the user. Use only when they ask to see, show, or view an artwork — not for list, summary, count, or text-only requests. Not for visual analysis by the LLM — use inspect_artwork_image to get image bytes. Not all artworks have images available. Returns metadata and a viewer link, not the image bytes themselves; do not construct or fetch IIIF image URLs manually (downloadable images are on rijksmuseum.nl).
Inputs: objectNumber (required). Returns: url, date, error, title, +8 more
get_conservation_historyGet Conservation HistoryARead-onlyIdempotent
Technical examinations and restoration history for ONE artwork. Follows get_artwork_details / a search result, by objectNumber. Returns technical examinations (X-ray, dendrochronology, paint samples, infrared), conservation/restoration treatment events, a count of recorded signature/inscription marks (use search_inscriptions for the actual transcriptions), and a short provenance excerpt. Not for general metadata — use get_artwork_details. Not for transcribed inscriptions — use search_inscriptions. Not for aggregate counts — use collection_stats.
Inputs: objectNumber (required). Returns: error, title, creator, warnings, +7 more
get_recent_changesGet Recent ChangesARead-onlyIdempotent
OAI-PMH delta feed of records changed within a date range. Paginated; anchored to a known harvest checkpoint. Use identifiersOnly=true for a lightweight listing (headers only, no full metadata). Full-mode records include an objectNumber for follow-up calls to get_artwork_details or get_artwork_image; identifiersOnly records carry only a LOD URI, so re-run without it to act on a change. Deleted records are flagged with deleted:true (marked [DELETED] in the listing) and carry only a LOD URI + datestamp, no metadata.
Inputs: from, until, setSpec, maxResults, +2 more. Returns: hint, error, records, totalChanges, +3 more
inspect_artwork_imageInspect Artwork ImageARead-onlyIdempotent
Returns image bytes (base64) for the LLM's own visual analysis. Covers a whole artwork or a region. The LLM can see and reason about the image immediately — not for the user to view (use get_artwork_image for the interactive viewer). Not for listing or summarising artworks — use search_artwork.
Use with region 'full' (default) to inspect the complete artwork, or specify a region to zoom into details, read inscriptions, or examine specific areas. The response includes cropPixelWidth/cropPixelHeight: the actual pixel dimensions of the returned image.
Region coordinates: 'pct:x,y,w,h' (percentage of full image, recommended), 'crop_pixels:x,y,w,h' (pixel coordinates of the full image — use with nativeWidth/nativeHeight from a prior response), or 'x,y,w,h' (legacy IIIF pixels, equivalent to crop_pixels). A tight region returns more real detail than a larger size — the server never upscales past the region's own pixels. For multi-panel works, estimate panel percentages from the physical dimensions in get_artwork_details.
When a viewer is open for this artwork, it automatically zooms to the inspected region (navigateViewer defaults to true, no effect when region is 'full'), so no separate navigate_viewer call is needed for basic zoom.
The response includes the active viewUUID (if any) for follow-up navigate_viewer calls.
Inputs: objectNumber (required), size, region, quality, +3 more. Returns: error, title, region, creator, +14 more
list_curated_setsList Curated SetsARead-onlyIdempotent
Browse thematic sets curated by Rijksmuseum staff. Includes sub-collection groupings (drawings, paintings, iconographic sets). Each result carries memberCount, top dominantTypes, top dominantCenturies by membership, and a category heuristic (object_type / iconographic / album / sub_collection / umbrella) so you can pick the right scope. Use minMembers: 100, maxMembers: 200000 to avoid umbrella sets when the user wants a substantive subset. Pair with browse_set(setSpec) to enumerate members. Not for keyword search across artworks — use search_artwork. Not for aggregate counts — use collection_stats.
Inputs: query, sortBy, maxMembers, minMembers, +1 more. Returns: sets, error, query, totalSets, +1 more
poll_viewer_commandsPoll Viewer CommandsC
Internal: poll for pending viewer navigation commands
Inputs: viewUUID (required). Returns: commands
remount_viewerRemount ViewerA
Internal: switch the viewer to a different artwork while preserving the viewUUID. Called by the artwork-viewer iframe during in-viewer related navigation. Any user highlight is cleared on remount because its coordinates belong to the previous artwork.
Inputs: viewUUID (required), objectNumber (required). Returns: url, date, error, title, +8 more
search_artworkSearch ArtworkARead-onlyIdempotent
Structured filter search — artworks matching ALL given filters. Filters cover subject, material, technique, date, place, person. Returns artwork summaries with titles, creators, and dates; every response includes totalResults (exact match count, not just the returned page). Not for free-text concept queries — use semantic_search for those. Not for artwork-to-artwork similarity — use find_similar with an objectNumber. For demographic person queries (gender, born/died, profession, birth/death place), use search_persons first to get a vocabId, then pass it as creator here. For provenance text and ownership history, use search_provenance. For aggregate counts and distributions, prefer collection_stats — one call vs compact=true loops.
Ranking: relevance (BM25) when text search (description, title, etc.) or geographic proximity is used; otherwise importance (image availability, curatorial attention, metadata richness). For concept-ranked results, use semantic_search.
At least one filter is required. There is no full-text search across all metadata. For concept or thematic searches (e.g. 'winter landscape', 'smell', 'crucifixion'), subject has the highest recall — it searches the large majority of the collection via structured Iconclass vocabulary. Use description for cataloguer observations (compositional details, specific motifs); use curatorialNarrative for curatorial interpretation and art-historical context. These three corpora can return complementary results. For broader concept discovery beyond structured vocabulary, use semantic_search — but combine it with search_artwork(type: 'painting', …) for painting queries since paintings are underrepresented there.
Array values are AND-combined (e.g. subject: ['landscape', 'seascape'] finds artworks with both). If many results share an object-number prefix (e.g. multiple folios of one sketchbook), a warnings note flags it; narrow with type/material filters or treat the shared prefix as the unit. Each result carries an objectNumber for follow-up calls to get_artwork_details (full metadata) or get_artwork_image (deep-zoom viewer for the user).
Inputs: sort, type, query, facets, +15 more. Returns: ids, error, facets, source, +4 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 13 tools
Tools are largely distinct and descriptions include explicit 'Not for X — use Y' clauses; the four viewer-related tools (get_artwork_image, inspect_artwork_image, navigate_viewer, poll_viewer_commands/remount_viewer) are clearly differentiated, with internal ones labeled. However, several descriptions redirect agents to tools that aren't in this set (semantic_search, search_persons, search_provenance, collection_stats, search_inscriptions), which creates ambiguous dead-end pointers.
Nearly all tools follow a verb_noun pattern (search_artwork, get_artwork_details, list_curated_sets, browse_set, inspect_artwork_image, navigate_viewer). Minor deviations exist: singular 'artwork' vs plural 'artworks' across search_artwork / find_artworks_citing_publication, and a couple of noun-first phrasings, but the set reads predictably.
13 tools is well within the sweet spot and each earns its place: core discovery, detail retrieval, image interaction, bibliography, conservation, curated sets, and delta sync are all represented without redundancy. Internal viewer tools are small helpers that don't bloat the surface.
Core lifecycle for browsing the collection (search → details → image → bibliography/conservation → sets → recent changes) is covered. But descriptions repeatedly point to missing tools (semantic_search for concept queries, search_persons for demographic vocabIds, collection_stats for aggregates, search_inscriptions), leaving genuine dead ends an agent cannot work around cleanly.
Maintenance
Related MCP Connectors
Rijksmuseum — the Dutch national museum's collection, via its open Data Services API…
Live Dutch open data for AI assistants: transport, weather, water, traffic, housing, government.
AI-native art catalogue. Catalogue works, parse provenance, and generate signed RAIs.
AI training dataset marketplace: 2M+ museum artworks with Golden Codex enrichment
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceAllows you to search for artworks, retrieve detailed information about specific artworks, access image tiles for artworks, and explore user-created collections from the Rijksmuseum.25 npm72MIT
- FlicenseNot gradedqualityDmaintenanceProvides access to European cultural heritage collections and artworks, allowing users to search for items, get detailed information about specific artworks, browse collections by institution, and receive AI-powered recommendations based on interests.1-
- AlicenseNot gradedqualityDmaintenanceEnables natural language search and exploration of the Dutch WWII Oorlogsbronnen archives, allowing users to query historical documents, photographs, and personal accounts through AI assistants.15MIT
- AlicenseNot gradedqualityCmaintenanceEnables searching the Rijksmuseum collection by keyword and retrieving full artwork details by ID.92 npmMIT