Knowledge Commons

One connector for the public knowledge commons.

Humanity's shared record — libraries, archives, historical newspapers, museums, scholarship, court opinions, government data — is publicly available but scattered across dozens of institutional APIs, each with its own conventions. Knowledge Commons federates them behind a single connector built on the Model Context Protocol, so any AI assistant can do research-grade work across all of them at once: reading a 1912 newspaper page, pulling a court opinion, crossing a census series with a scholarly citation graph.

Rather than exposing hundreds of endpoints, the connector offers five meta-tools and lets the model discover the rest at runtime:

list_sources — the catalog of everything the connector can reach

search_tools — semantic search over every capability, including a 12,000-server index of the community MCP registry

call_tool — execute any tool on any source

resolve_entity — resolve a name to a Wikidata QID with identifier crosswalk (VIAF, LCCN, ORCID) so results join across archives

deep_research — an archival research playbook the model executes step by step

Sources

Internet Archive — 40M+ digitized texts, audio, and film

Library of Congress — Collections plus Chronicling America newspapers, 1770–1963, full text

DPLA — 50M+ items from US libraries, archives, and museums

Open Library — Every published book: editions, authors, full-text links

Wikipedia / Wikidata / Commons — Encyclopedia, knowledge graph with SPARQL, 100M+ free images

OpenAlex — 250M+ scholarly works with citation graph

arXiv / PubMed — Preprints plus 37M biomedical citations

CourtListener — US court opinions, full-text searchable

SEC EDGAR — Every US public-company filing since 2001, full text

Census / FRED / World Bank — Demographics plus 800K+ economic time series

GDELT — Global news in 100+ languages, updated every 15 minutes

The Met / Smithsonian / Europeana — Open-access museum collections, US and Europe

openFDA / ClinicalTrials.gov — Drug labels, adverse events, recalls, and clinical trials

Federal Register / GBIF / USGS / NASA / Open-Meteo — US rulemaking, biodiversity, earthquakes, space imagery, weather

Every result carries a citation back to the holding institution. Entity resolution runs on a Wikidata spine, so a person, place, or work found in one archive can be followed into the others.

Try it

It speaks the Model Context Protocol over streamable HTTP, so any MCP client can use it. In Claude, open Settings → Connectors → Add custom connector and paste:

https://vps.benelias.moe/commons/90AXhiNjAYEBGj4sq0NH4mMJ/mcp

This is a shared demo endpoint on one small server, so it is rate limited and may be slow when several people are using it. Every source listed above is available on it. The one capability it withholds is proxying to community MCP servers from the public registry — the demo host does not relay to third parties on a caller's behalf.

To run your own copy, with community-registry proxying and your own keys, the source is on GitHub.

Contributing a source is a small piece of work: one file holding a description the model reads and a few tools wrapping a public API. For that, or anything else, write me at beneliasmoe [at] gmail dot com.