uc-catalog-mcp

MCP server for the University of Chicago Library catalog (VuFind Search & Record API). Exposes two tools over stdio: search and get_record.

Catalog access / VPN

The Search & Record API is typically reachable only from campus network or University VPN. Off-network callers may receive a bot-check page or HTTP 403 instead of JSON. This server fails soft in those cases and does not invent holdings.

Override the API base with UC_CATALOG_BASE (default https://catalog.lib.uchicago.edu/vufind).

Install

Clone this repository, then install dependencies into that checkout (editable install is enough for local MCP use):

cd /path/to/uc-catalog-mcp
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Requires uv on your PATH for the Claude config below (uv run uses the project environment).

Claude / MCP host config

Point Claude Desktop (or another stdio MCP host) at the checkout with uv run --directory. Replace /path/to/uc-catalog-mcp with your clone path:

{
  "mcpServers": {
    "uc-catalog": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/path/to/uc-catalog-mcp",
        "python",
        "-m",
        "uc_catalog_mcp"
      ],
      "env": {
        "UC_CATALOG_BASE": "https://catalog.lib.uchicago.edu/vufind"
      }
    }
  }
}

Do not rely on "command": "python" plus "cwd" alone — that form often fails to resolve the package in Claude’s MCP launcher. uv run --directory keeps the working tree and environment explicit.

After an editable install you can also run the console script uc-catalog-mcp from a shell for a quick smoke check.

Tools

search

Argument	Notes
lookfor	Query string (required)
type	AllFields (default), Title, Author, Subject, ISN, …
filter	Repeatable VuFind filters, e.g. format:"Book", building:"Regenstein Library"
facet	Repeatable facet fields, e.g. format, building
limit	Page size (default 10, max 100)
page	Result page (default 1)
sort	Optional VuFind sort

Returns resultCount, records (id, title, authors, formats, buildings, callNumbers when present, record permalink), facets when requested, and a human catalog searchUrl for the same query.

buildings comes from the Solr building field; callNumbers from callnumber-raw (indexed holdings metadata, not live circulation).

get_record

Argument	Notes
id	Single catalog record id (required; no batch)

Same core fields as search records, plus permalink / buildings / callNumbers.

What this does not do

• Enrichment (WikiData, Internet Archive, PubMed, HathiTrust, …)
• Live FOLIO availability / circulation status
• MyAccount, holds, or patron actions
• A separate list_facets tool (request facets via search)

Development

pytest

Unit tests mock HTTP. Live checks against the catalog require campus network or University VPN.

License

MIT