About Provenio — Art Intelligence
Provenio is a hosted art-provenance and market-intelligence service for AI agents and researchers. Its cited records cover transaction history, custody chains, risk windows, and explicit gaps in the available evidence.
Key features
- 01Search cited provenance records
- 02Trace artwork custody chains
- 03Find comparable artworks and auction prices
- 04Explore influence networks and reception
- 05Flag historical risk periods
Use cases
- 01Research an artwork’s ownership history
- 02Compare auction results for market research
- 03Check records for custody gaps and risk flags
MCP support
Provenio provides a hosted MCP server at https://provenio.art/api/mcp, exposing tools and prompts for art provenance and market research. It supports streamable HTTP and does not require an authorization header.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://provenio.art/api/mcpclaude mcp add --transport http mildo-ai-provenio https://provenio.art/api/mcp{"mcpServers":{"mildo-ai-provenio":{"type":"http","url":"https://provenio.art/api/mcp"}}}Open in Cursor{"servers":{"mildo-ai-provenio":{"type":"http","url":"https://provenio.art/api/mcp"}}}Open in VS Code{"mcpServers":{"mildo-ai-provenio":{"type":"http","url":"https://provenio.art/api/mcp"}}}{"mcpServers":{"mildo-ai-provenio":{"serverUrl":"https://provenio.art/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://provenio.art/api/mcpAuth requiredTools
Details
From the project READMEProvenio — Art Provenance MCP
Provenance and market intelligence for AI agents. A 282,731-node knowledge graph over 273,378 art-market transactions — the Getty Provenance Index, the Knoedler stock books (1872–1971), ERR and wartime-restitution records — served to any MCP-compliant client.
Every answer is cited. Custody gaps are named, never guessed. Hallucination is held under 1%, verified by an externally reproducible evaluation harness rather than self-report. Where the record is silent, Provenio returns an explicit gap — it never asserts clean title.
🌐 https://provenio.art · Free open beta — no key, no card.
Quick start
Add the remote server to any MCP client:
https://provenio.art/api/mcp
- Claude (claude.ai / Desktop): Settings → Connectors → Add custom connector → paste the URL
- ChatGPT (MCP enabled): add as a connector —
search/fetchcompatible - Cursor / other clients: streamable-http transport, no auth header needed
Open beta: 5,000 queries/month per IP, all 27 tools, full corpus. Need a higher cap? Write to [email protected].
What's inside
| Axis | Coverage |
|---|---|
| Knowledge graph | 282,731 nodes · CIDOC-CRM / PROV-O aligned |
| Market transactions | 273,378 (99,892 with verified inflation-adjusted prices) |
| Provenance corpus | Getty Provenance Index · Knoedler 1872–1971 · ERR · curated wartime restitution |
| Risk flagging | Nazi-era 1933–1945 (HEAR Act) · colonial-era · dealer-fraud windows |
| Non-Western depth | Korean Dansaekhwa (Kim Whanki, Lee Ufan, Park Seo-Bo) market + exhibition histories |
| Tools | 27 typed tools · 15 prompts — search, provenance chains, comparables, influence networks, reception arcs, BYOK federation |
Tools (selection)
get_provenance_chain— full custody chain with verification state and risk-window flags; always returnschecked_databasessearch_auction_history— lot-level prices with optional P25/50/75/90 percentile summary in one callfind_comparable_artworks— era-precise, market-guarded comparablesget_influence_network/get_reception_arc— scholarly influence and verbatim critical reception with source refsbyok_external_query— federate your own Artnet/Artprice/gallery-DB credentials (SSRF-guarded)
Full catalogue: provenio.art/docs · OpenAPI 3.1 spec
Registries
- Official MCP Registry:
io.github.mildo-ai/provenio - Smithery:
mildo-ai/provenio
Honesty engineering
Provenio's design contract, in one line: no model memorises art history —
Provenio lets any model cite it. Every response binds to a verifiable source
row; absences come back as named coverage_gaps; the evaluation harness that
certifies the <1% hallucination rate is re-runnable against the live API.
Read the methodology: The Invisible Gaps · Genesis — the artwork
Feedback & data corrections
Found a wrong price, a broken custody chain, a missing attribution — or want a tool that doesn't exist yet? Open an issue. Data corrections are applied only with a citable source, and every applied correction is noted in the release log. You can also write to [email protected].
Built and operated by Hoyeob Kim · Mildo Inc. · Contact: [email protected]
This repository hosts the public server manifest and documentation for the hosted MCP service. The knowledge-graph pipeline and web application live in a private monorepo.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.