@cyanheads/libofcongress-mcp-server

Public Hosted Server: https://libofcongress.caseyjhand.com/mcp

Tools

Six tools covering the Library of Congress digital holdings — general search with format/date/subject/location filters, full item retrieval, Chronicling America newspaper search with OCR, single-page full-text fetch, LCSH subject heading lookup, and curated collection browsing:

libofcongress_search

Search the LOC digital collections with full-text keyword matching and facet filters.

• Eight material formats: photo, map, newspaper, manuscript, audio, film, book, notated-music

• Date range filtering by year (inclusive start and end)

• Subject heading filter — use libofcongress_search_subjects first to get the exact LCSH spelling

• Geographic location filter (e.g., "oklahoma", "washington d.c.")

• collection_slug scopes the search to one curated collection — pass a slug from libofcongress_browse_collections. Mutually exclusive with format (each selects a different LOC endpoint); an unknown slug returns collection_not_found

• Pagination up to 100 results per page — pages/has_next respect LOC's ~100,000-item retrieval ceiling (a notice discloses how to reach the rest: partition by date, subject, or location); real results on a page beyond the reported total are always returned, never discarded

• Empty results include a notice field with recovery hints — echoes the applied filters

• Each result carries is_item — true for catalog items whose id resolves via libofcongress_get_item, false for non-item results (collections, exhibit/guide pages, newspaper pages); open their url instead

libofcongress_get_item

Retrieve the full metadata record for a specific LOC digital item.

• Returns contributors, LCSH subject headings, rights information, physical/technical description, and cataloger notes

• Also returns summary, languages, locations, call_number (shelf location for requesting the physical original), former_ids, original_formats, online_formats, and access_restricted — all sourced from the same upstream response, no extra request

• resource_links contains URLs to downloadable digital files (TIFF, JPEG, PDF) for items with digital surrogates

• related_items lists IDs of related LOC items for follow-up retrieval

• resource_links and related_items render in full in both structuredContent and content[] — no truncation, so content[]-only clients see every value

• Deduplicates resource links from nested files[] arrays

• Accepts multi-segment item IDs verbatim (e.g. newspaper pages sn95047246/1935-09-05/ed-1); returned url is always an absolute https:// URL

• Fields absent upstream are omitted rather than filled — a sparse record stays sparse

libofcongress_search_newspapers

Search historical newspaper pages in the Chronicling America corpus via the LOC /newspapers/ endpoint.

• OCR text excerpts (~500 chars) returned inline for relevance assessment without a second hop

• Filters: keyword, date range, US state (full state name), newspaper publication title (partial match)

• Returns the url field needed by libofcongress_get_newspaper_page — do not construct these URLs manually

• OCR quality varies by digitization batch and era; 19th-century and degraded materials may contain garbled text

• Empty results include a notice with recovery suggestions (broaden date, try different keywords, historical OCR caveat)

libofcongress_get_newspaper_page

Retrieve the full OCR text and metadata for a specific newspaper page.

• Accepts the url field from a libofcongress_search_newspapers result — validates the URL prefix before any outbound request

• Fetches JSON from the LOC text-services endpoint (tile.loc.gov) and reads plain text from the full_text field

• ocr_available: false when the page has no digitized text (image-only batch) — not an error, a data property

• When ocr_available is true but the text service returns nothing, a notice discloses the retrieval miss on both response surfaces — distinct from a genuinely image-only page

• Strips echoed q= params from fulltext URLs to avoid tile.loc.gov 404s (known LOC API quirk)

libofcongress_search_subjects

Search Library of Congress Subject Headings (LCSH) via id.loc.gov.

• Returns standardized labels and stable LOC URIs for subjects matching the keyword

• count field indicates approximate number of LOC items carrying that heading (when available)

• Use the returned label exactly in the libofcongress_search subject filter — LCSH uses inverted forms ("Photography, Aerial", "World War, 1939-1945") that differ from natural language

• Draws from the id.loc.gov suggest endpoint's full 50-candidate pool (not scaled to limit) and filters to true LCSH headings, so a heading ranked below name-authority records isn't reported as a false empty; when that ranked pool — rather than a lack of coverage — yields an empty or short result, the response discloses it with a recovery hint

libofcongress_browse_collections

List and browse LOC curated digital collections.

• Returns collection slug — pass it to libofcongress_search as collection_slug to search inside that collection

• Slugs come from the collection's loc.gov route, not its title — "Aaron Copland Collection" lives at aaron-copland, so they aren't guessable

• Optional keyword filter by collection name/description

• Item counts are approximate; omitted when the API doesn't provide them

• Pagination supported up to 100 collections per page

Related MCP server: historical-investigator-mcp

Resource

All resource data is also reachable via libofcongress_get_item. Use libofcongress_search to discover item IDs first.

Write the item ID with its slashes intact — libofcongress://item/sn95047246/1935-09-05/ed-1 is the canonical form for a multi-segment newspaper ID. Percent-encoded slashes (%2F) also resolve.

Features

Built on @cyanheads/mcp-ts-core:

• Declarative tool and resource definitions — single file per primitive, framework handles registration and validation

• Unified error handling — handlers throw, framework catches, classifies, and formats

• Pluggable auth: none, jwt, oauth

• Swappable storage backends: in-memory, filesystem, Supabase, Cloudflare KV/R2/D1

• Structured logging with optional OpenTelemetry tracing

• STDIO and Streamable HTTP transports

LOC-specific:

• Module-level rate-limit enforcement: 20 req/min limit; 429 responses trigger a 1-hour block with per-minute countdown in error messages

• Configurable pacing delay (default 3100ms, ~19 req/min) applied before every outbound LOC API request

• HTML-response detection guards against silent rate-limit proxy pages that return 200 with HTML

• Out-of-range page handling: LOC returns HTTP 400 or 520 for page numbers beyond the result set — treated as empty rather than errors

• Transient-fault resilience: network drops and timeouts retry with backoff behind a 30s per-request timeout ceiling; the 429 rate-limit path is never retried, since a retry would deepen LOC's 1-hour block

• JSON OCR extraction for newspaper text — reads the full_text field from LOC text-services responses

• Two-service architecture: LocApiService for www.loc.gov and LcLinkedDataService for id.loc.gov

Agent-friendly output:

• Empty results always include a notice field with recovery hints — echoes the applied filters and suggests how to broaden

• Pagination status on every search response: total, page, pages, has_next — capped at LOC's ~100,000-item retrieval ceiling, with a notice disclosing how to page past it

• ocr_available discriminator on newspaper page results so callers can branch on data availability without parsing text

• Recovery hints on all error contracts — actionable next steps for the agent on every failure mode

Getting started

Add the following to your MCP client configuration file.

{
  "mcpServers": {
    "libofcongress-mcp-server": {
      "type": "stdio",
      "command": "bunx",
      "args": ["@cyanheads/libofcongress-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_SESSION_MODE": "stateless",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

Or with npx (no Bun required):

{
  "mcpServers": {
    "libofcongress-mcp-server": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cyanheads/libofcongress-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_SESSION_MODE": "stateless",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

Or with Docker:

{
  "mcpServers": {
    "libofcongress-mcp-server": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "MCP_TRANSPORT_TYPE=stdio",
        "-e", "MCP_SESSION_MODE=stateless",
        "ghcr.io/cyanheads/libofcongress-mcp-server:latest"
      ]
    }
  }
}

For Streamable HTTP, set the transport and start the server:

MCP_TRANSPORT_TYPE=http MCP_HTTP_PORT=3010 bun run start:http
# Server listens at http://localhost:3010/mcp

Prerequisites

• Bun v1.3.0 or higher (or Node.js v24+).

• No API key required — the LOC JSON API and LC Linked Data endpoints are open. LOC recommends a descriptive LOC_USER_AGENT for polite access.

Installation

1. Clone the repository:

git clone https://github.com/cyanheads/libofcongress-mcp-server.git

2. Navigate into the directory:

cd libofcongress-mcp-server

3. Install dependencies:

bun install

4. Configure environment:

cp .env.example .env
# edit .env if you want to set LOC_USER_AGENT or LOC_REQUEST_DELAY_MS

Configuration

All configuration is validated at startup via Zod schemas in src/config/server-config.ts.

See .env.example for the full list of optional overrides.

Running the server

Local development

• Build and run:

  # One-time build
  bun run rebuild
  
  # Run the built server
  bun run start:stdio
  # or
  bun run start:http

• Run checks and tests:

  bun run devcheck   # Lint, format, typecheck, security
  bun run test       # Vitest test suite
  bun run lint:mcp   # Validate MCP definitions against spec

Docker

docker build -t libofcongress-mcp-server .
docker run --rm -p 3010:3010 libofcongress-mcp-server

The Dockerfile defaults to HTTP transport and logs to /var/log/libofcongress-mcp-server.

Project structure

Development guide

See CLAUDE.md for development guidelines and architectural rules. The short version:

• Handlers throw, framework catches — no try/catch in tool logic

• Use ctx.log for request-scoped logging, ctx.state for tenant-scoped storage

• Register new tools and resources via the arrays in src/index.ts

• Wrap external API calls: validate raw → normalize to domain type → return output schema; never fabricate missing fields

Contributing

Issues and pull requests are welcome. Run checks and tests before submitting:

bun run devcheck
bun run test

License

Apache-2.0 — see LICENSE for details.