Entscheidsuche — Schweizer Gerichtsentscheide

What Entscheidsuche — Schweizer Gerichtsentscheide does

Entscheidsuche — Schweizer Gerichtsentscheide provides MCP access to published Swiss court decisions and related legal documents. Its indexed collection covers federal courts, courts in all 26 cantons, administrative authorities, and penal orders. Documents are available in German, French, and Italian, with language filtering and cross-language search supported by the search interface.

The public service is operated by the non-profit association entscheidsuche.ch. Its data collection is built from regular scrapes of original court sources and includes material submitted through the project’s upload function. The repository describes the service as beta software.

How it works

The hosted endpoint is https://mcp.entscheidsuche.ch/mcp. It uses Streamable HTTP and does not require authentication or an API key. MCP clients can connect directly using a URL; clients limited to stdio can use the mcp-remote bridge. The repository also supports running the application locally over Streamable HTTP or stdio.

Search uses Lucene syntax. Queries can combine terms, phrases, OR expressions, negation, wildcards, and field-specific searches. Results can be restricted by decision date, scrape date, language, canton, court, or chamber, then sorted by relevance, decision date, or scrape date. Pagination uses next_cursor and search_after, while aggregations support facet-oriented result analysis.

Tools and capabilities

The Entscheidsuche — Schweizer Gerichtsentscheide tool set includes:

• search for full-text queries, filters, sorting, pagination, and aggregations.
• search_by_case_number for exact phrase searches covering file numbers, docket references, and BGE citations. The tool adds quotation marks automatically.
• fetch_document for retrieving a complete decision and its full text by document ID.
• list_hierarchy for finding federal, cantonal, court, and chamber hierarchy IDs together with result counts.
• list_facets for reading a hierarchical facet tree with localized German, French, and Italian labels.
• server_info for version and configuration details.

A typical workflow is to discover relevant material with search, narrow or verify a citation with search_by_case_number, and then retrieve the complete text using fetch_document. list_hierarchy and list_facets help clients build structured filters before issuing a search.

Setup and configuration

For a hosted connection, add the endpoint URL to an MCP client. Claude Code can register it with an HTTP transport. Cursor, Windsurf, and VS Code can use an mcp.json entry containing the server URL. Claude Desktop requires the mcp-remote bridge because its documented setup is stdio-only. The MCP Inspector can also connect to the endpoint for testing.

Local development requires Python, a virtual environment, and installation of the package with its development extras. The module starts an HTTP server on 127.0.0.1:8765/mcp by default, or it can be started with the stdio transport. Local configuration can set the Elasticsearch endpoint, facet source, host, port, MCP path, stateless HTTP behavior, DNS rebinding protection, public base URL, and log level. Each has a documented default in the repository.

Limitations and notes

The repository labels the server beta and notes that client-specific instructions, tool lists, pricing requirements, and configuration paths may change. The public endpoint currently has no authentication. The available records are published or collected legal documents; the material does not describe the service as a legal advice system. Search results without an explicit language filter return the first available language field in the order German, French, then Italian.