🇨🇭 Part of the Swiss Public Data MCP Portfolio

📚 swiss-academic-libraries-mcp

MCP server providing access to Swiss academic libraries — swisscovery, e-rara, e-periodica, e-manuscripta. No API key required.

🇩🇪 Deutsche Version

Demo

Overview

swiss-academic-libraries-mcp connects AI models to the full Swiss academic library infrastructure via standardised, open protocols. It covers the swisscovery union catalogue (500+ libraries, 10M+ records) and three digitalisation platforms: historical prints (e-rara), periodicals (e-periodica) and manuscripts (e-manuscripta).

All data sources use open, authentication-free protocols (SRU/MARC21, OAI-PMH/Dublin Core). The server supports both local use via Claude Desktop (stdio transport) and cloud deployment (Streamable HTTP).

Anchor demo query: "Which Swiss university dissertations on primary school pedagogy are held in Swiss libraries, and are any of them digitised in e-rara?"

Features

• 11 tools across 4 data sources — all read-only, no API key required

• swisscovery search with full CQL syntax: full-text, title, author, subject, ISBN/ISSN

• OAI-PMH harvesting with date range and collection filters plus pagination via resumption tokens

• MARC21 parser extracting 20+ fields (title, creator, publication info, subjects, abstract, URLs)

• Dublin Core parser for all three digitalisation portals

• Dual transport: stdio for Claude Desktop · Streamable HTTP for cloud/self-hosted deployments

• 2 built-in prompts: research-workflow and education-research

• Markdown and JSON output for all tools

• 34 unit tests (no network) + 6 live smoke tests

Data Sources

Tools

Example Use Cases

Prerequisites

• Python 3.11 or higher

• uv / uvx (recommended) or pip

• Internet access (all APIs are publicly available)

Installation

Claude Desktop (recommended)

Add to claude_desktop_config.json:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "swiss-academic-libraries": {
      "command": "uvx",
      "args": ["swiss-academic-libraries-mcp"]
    }
  }
}

Restart Claude Desktop — the server starts automatically on first use.

Cloud / Self-hosted (Streamable HTTP)

uvx swiss-academic-libraries-mcp --http --port 8000 [--host 127.0.0.1]

Security & Deployment Notes

• Default binding is 127.0.0.1 (loopback only). The server has no built-in authentication.

• Use --host 0.0.0.0 only when running behind a reverse proxy that provides authentication and per-IP rate limits (e.g. nginx with limit_req + OAuth2-Proxy). Non-loopback bindings emit a WARN log.

• Logs go to stderr; set verbosity with MCP_LOG_LEVEL=DEBUG|INFO|WARNING.

Development

git clone https://github.com/malkreide/swiss-academic-libraries-mcp
cd swiss-academic-libraries-mcp
pip install -e .

Quickstart

Start by calling library_info for a full overview. Then:

"Which books about Swiss primary schools are held in Swiss libraries?"
→ swisscovery_search(query='subject = "Volksschule"', max_records=20)

"Show digitised historical works from ETH Library"
→ erara_list_records(set_spec="zut")

"Which Swiss periodicals were digitised in 2023?"
→ eperiodica_list_records(from_date="2023-01-01", until_date="2023-12-31")

"What manuscript collections does e-manuscripta hold?"
→ emanuscripta_list_collections()

→ [More use cases by audience](EXAMPLES.md) →

💡 "No API key — just install and query."

CQL Search Syntax (swisscovery)

Full text:     Volksschule Zürich
Title:         title = "education reform"
Author:        creator = "Pestalozzi"
Subject:       subject = "pedagogy"
ISBN:          isbn = "978-3-05-006234-0"
Combined:      title = "school" AND creator = "Pestalozzi"
Pagination:    start_record = 11

Configuration

No API keys or environment variables required.

Project Structure

swiss-academic-libraries-mcp/
├── src/
│   └── swiss_academic_libraries_mcp/
│       ├── __init__.py       # Package init
│       ├── server.py         # FastMCP server, 11 tools, 2 prompts, 1 resource
│       └── api_client.py     # HTTP client, MARC21 + OAI-PMH/DC parsers
├── tests/
│   └── test_server.py        # 34 unit tests + 6 live smoke tests
├── pyproject.toml
├── CHANGELOG.md
├── CONTRIBUTING.md
├── LICENSE
├── README.md                 # This file (English)
└── README.de.md              # German version

Testing

# Unit tests (no network required)
PYTHONPATH=src pytest tests/ -m "not live"

# Live smoke tests (internet required)
PYTHONPATH=src pytest tests/ -m "live"

Safety & Limits

• Read-only: All tools perform HTTP GET requests against public SRU and OAI-PMH endpoints — no data is written, modified, or deleted.

• No personal data: The APIs return bibliographic metadata (titles, authors, publication info, subject headings) and public digitisation records. No personally identifiable information (PII) about library users is processed or stored.

• Rate limits: swisscovery SRU and the OAI-PMH endpoints are public and have no documented hard limits, but OAI-PMH harvesting is paginated via resumption tokens — use from_date / until_date and keep max_records reasonable. The server enforces a 30s timeout per request.

• Data freshness: Results reflect the upstream catalogues at query time. No caching is performed by this server; indexing latency is controlled by SLSP and the digitisation platforms.

• Terms of service: Data is subject to the ToS and licences of each source — swisscovery / SLSP, e-rara, e-periodica, e-manuscripta. Most digitised material is in the public domain or under Creative Commons licences; always check the rights statement on the individual record before redistribution.

• No guarantees: This is a community project, not affiliated with SLSP, ETH Library, or any of the participating institutions. Availability depends on the upstream APIs.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines on:

• Reporting bugs and requesting features

• Setting up the development environment

• Code style and test requirements

• Submitting pull requests

This project follows the conventions of the Swiss Public Data MCP Portfolio.

Changelog

See CHANGELOG.md

Deployment for Swiss Public Administration

If you self-host this server for a Swiss school authority, archive, or municipal use case:

• Data residency: prefer on-premise or a CH-based cloud provider. The query patterns themselves (which library searches a civil servant runs) may reveal ongoing research and are best kept on Swiss infrastructure.

• Upstream calls go exclusively to CH-hosted services: SLSP / swisscovery, ETH-Bibliothek (e-rara, e-periodica, e-manuscripta). No data leaves Switzerland.

• Logging: logs are written to stderr; configure your IT retention policy accordingly (e.g. systemd-journal MaxRetentionSec).

• HTTP transport must run behind a reverse proxy with authentication and per-IP rate limits (see Security & Deployment Notes above).

License

MIT License — see LICENSE

Author

Hayal Oezkan · github.com/malkreide

Credits & Related Projects

• Data: swisscovery / SLSP · e-rara · e-periodica · e-manuscripta

• Protocol: Model Context Protocol — Anthropic / Linux Foundation

• Related: eth-library-mcp — ETH Library Discovery & Persons API

• Portfolio: Swiss Public Data MCP Portfolio