MCP Libraryofcongress

What the mcp-libraryofcongress MCP server does

The mcp-libraryofcongress MCP server gives an AI client access to selected Library of Congress discovery functions. It can search the words found inside scanned US newspaper pages, search catalogue records, retrieve an individual item record, and list digital collections published by Library of Congress curators.

Newspaper searching uses text produced by optical character recognition. Results can include the matching page’s identifier, title, contributing library, year, page number, publication details, state, excerpts, and a source URL. Catalogue searches can cover books, photographs, maps, audio, manuscripts, or newspapers, with filters such as date, subject, location, language, and collection.

How it works

The server communicates with an MCP client over the standard local process configuration. A client invokes one of four tools:

• search_newspapers searches OCR text from scanned newspaper pages.
• search_items searches Library of Congress catalogue records.
• get_item reads a record, its rights information, and where the original is held.
• list_collections returns the available digital collections.

A typical research flow starts with search_newspapers or search_items. Search results provide an identifier, which can then be passed to get_item for more record detail. Newspaper results are paginated by matching leaves, not by the number of times a phrase appears. Excerpts may be centered on a match or may show the beginning of a page when the returned text does not include the matching words, so excerpts should be interpreted alongside their metadata.

Setup and configuration

The mcp-libraryofcongress MCP server does not require an API key, account, environment variable, volume, port, or credential. The npm-based setup requires Node 24 or later:

{
  "mcpServers": {
    "libraryofcongress": {
      "command": "npx",
      "args": ["-y", "mcp-libraryofcongress"]
    }
  }
}

Claude Code can add it with claude mcp add libraryofcongress -- npx -y mcp-libraryofcongress. Docker is also supported with the image ghcr.io/smeet666/mcp-libraryofcongress:3.0.2; stdin must remain open, and the container needs outbound HTTPS access to www.loc.gov and chroniclingamerica.loc.gov. An .mcpb bundle is available for clients that support MCP bundles and includes its dependencies.

Tools and capabilities

search_newspapers requires a query of 1–300 characters. Optional filters include publication location, publication name, year range, result limit, page number, excerpt length, and excerpts per match. Its result limit defaults to 10 and can be set from 1 to 25.

search_items requires a query of 1–300 characters. It supports media-type, year, subject, location, language, and collection filters. get_item uses an item identifier from a search result. list_collections lists the digital collections exposed by the Library of Congress.

Limitations and notes

Newspaper matches depend on OCR text, so the returned text can differ from the scan. A page_opening excerpt does not necessarily contain the query match; it is provided when the Library of Congress response does not expose the matching text in that excerpt. The total value for newspaper searches counts matching newspaper leaves, not occurrences of the query.

The mcp-libraryofcongress MCP server is focused on discovery and record retrieval.