fedlex-mcp

MCP server for Swiss federal legislation via Fedlex.

Provides two complementary layers of access:

• Live SPARQL tools — query the Fedlex SPARQL endpoint in real-time (no local setup beyond npm run build)
• SQLite-backed tools — fast, offline-capable tools that work against a local database built by npm run ingest

Experimental project. Not legal advice. Always verify against the official Fedlex source.

---

Tools

Live SPARQL tools

These query https://fedlex.data.admin.ch/sparqlendpoint directly and require no local database.

Tool	Description
search_legislation	Search by keyword in law titles and abbreviations
list_legislation	List all in-force legislation (paginated)
get_legislation	Get metadata for a law by SR number or ELI URI
get_legislation_text	Fetch the full text of a law (HTML or XML)
get_versions	List all historical consolidation versions
get_citations	Law-to-law citation graph (which laws cite which)

SQLite-backed tools

These work against a local SQLite database built by npm run ingest. They are faster, work offline, and provide capabilities not available via SPARQL alone.

Tool	Description
search_provisions	FTS5 full-text search across all article texts (~100k provisions)
lookup_definition	Find formal legal definitions by term across all legislation
find_eu_references	Swiss laws that implement or reference specific EU directives/regulations
get_article_citations	Article-to-article citation graph (which specific articles cite which)
get_db_stats	Database statistics: counts, size, and last ingestion timestamp

get_article_citations is more precise than get_citations: instead of "Law A cites Law B", it returns "Art. 12 of the Criminal Code cites Art. 3 of the Data Protection Act".

---

Identifiers

Every tool that takes an identifier accepts either:

• An SR number — e.g. "101" (Federal Constitution), "210" (Civil Code), "311.0" (Criminal Code), "220" (Code of Obligations)
• A full ELI URI — e.g. "https://fedlex.data.admin.ch/eli/cc/1999/404"

Languages

All tools support language: "fr" | "de" | "it" | "rm" (default: "fr").

---

Setup

1. Install and build

npm install && npm run build

2. (Optional) Build the local SQLite database

Required for the SQLite-backed tools. Downloads and parses all HTML legislation from Fedlex (~7,000 laws, typically 15–30 minutes depending on connection):

npm run ingest

Options:

Flag	Default	Description
--lang	fr	Language to ingest (de, fr, it)
--concurrency	5	Parallel HTML fetches
--db	./data/fedlex.sqlite	Output path for the SQLite file

You can also set FEDLEX_DB_PATH as an environment variable to point the server at an existing database.

The ingestion pipeline:

1. Fetches all ConsolidationAbstract entries from SPARQL
2. Resolves the latest consolidation and download URLs per law
3. Fetches and parses HTML documents to extract article provisions
4. Extracts legal definitions and EU cross-references per article
5. Rebuilds the FTS5 full-text index
6. Fetches article-level citation edges from SPARQL and stores them
7. Writes a JSON summary report alongside the database file

---

Usage

stdio (local / Claude Desktop)

node dist/index.js

Claude Desktop config (~/.claude/claude_desktop_config.json):

{
  "mcpServers": {
    "fedlex": {
      "command": "node",
      "args": ["/path/to/fedlex-mcp/dist/index.js"]
    }
  }
}

To use a pre-built database, add FEDLEX_DB_PATH to the env:

{
  "mcpServers": {
    "fedlex": {
      "command": "node",
      "args": ["/path/to/fedlex-mcp/dist/index.js"],
      "env": { "FEDLEX_DB_PATH": "/path/to/fedlex.sqlite" }
    }
  }
}

HTTP/SSE (online deployment)

node dist/index.js --http
# custom port:
PORT=8080 node dist/index.js --http

MCP endpoints:

• GET /sse — SSE stream for MCP clients
• POST /messages?sessionId=<id> — message relay

Docker

docker build -t fedlex-mcp .
docker run -p 3000:3000 fedlex-mcp

Deploy to Railway / Render / Fly.io

Set the start command to node dist/index.js --http and expose port 3000 (or set PORT env var).

---

REST API

All tools are also available as plain HTTP GET endpoints under /api/. Useful for browser clients or non-MCP integrations.

Live SPARQL endpoints

GET /api/search?q=<query>&lang=fr&limit=20&in_force=true
GET /api/list?lang=fr&limit=50&offset=0&in_force=true
GET /api/legislation/:id?lang=fr
GET /api/text/:id?lang=fr&format=html&version=2024-01-01
GET /api/versions/:id
GET /api/citations/:id?direction=both&lang=fr&limit=50

SQLite-backed endpoints

GET /api/db/search?q=<query>&limit=20&sr=235.1
GET /api/db/definitions?term=<term>&limit=20
GET /api/db/eu-refs?id=2016/679&type=Directive&limit=30
GET /api/db/article-citations?sr=311.0&article=art_1&direction=both&limit=50
GET /api/db/stats

Parameters for /api/db/article-citations:

Parameter	Required	Description
sr	Yes	SR number of the law
article	No	Filter to a specific article by HTML id (e.g. art_1, art_2a)
direction	No	from (outgoing), to (incoming), or both (default)
limit	No	Max results per direction (default 50, max 200)

---

Examples

Live SPARQL tools

search_legislation("protection des données")
→ finds LPD (SR 235.1) and related ordinances

get_legislation("235.1")
→ metadata for the Loi fédérale sur la protection des données

get_legislation_text("311.0", language="fr")
→ full text of the Code pénal suisse

get_versions("220")
→ all historical versions of the Code des obligations

get_citations("101", direction="from", language="fr")
→ all laws that cite the Constitution fédérale

search_legislation("Datenschutz", language="de")
→ same search in German

SQLite-backed tools

search_provisions("données personnelles sensibles", sr_number="235.1")
→ FTS5 search within the LPD for that phrase

lookup_definition("personne concernée")
→ all formal definitions of that term across Swiss federal law

find_eu_references(eu_identifier="2016/679")
→ Swiss laws referencing the GDPR

get_article_citations("311.0", article_id="art_12", direction="both")
→ articles that Art. 12 CP cites, and articles in other laws that cite Art. 12 CP

get_db_stats()
→ total counts of laws, provisions, definitions, EU references, article citations

---

Data source

All data comes from the Swiss Federal Chancellery's Fedlex platform via:

• SPARQL endpoint: https://fedlex.data.admin.ch/sparqlendpoint
• Ontology: JOLux (based on FRBR, ELI)
• Formats available: PDF, HTML, XML (Akoma Ntoso, since 2022), DOCX
• Coverage: All Classified Compilation (RS/SR) texts — acts, ordinances, regulations, treaties

---

Disclaimer

This project is an independent, experimental MCP server for accessing information derived from the Swiss Federal Chancellery's Fedlex platform. It is provided for technical and informational purposes only.

• No legal advice. Nothing produced by this project constitutes legal advice, legal opinion, or a substitute for professional counsel.
• No warranty. This software is provided "as is", without warranty of any kind. Outputs may be incomplete, inaccurate, unavailable, improperly parsed, or out of date.
• Verify with official sources. Users must independently verify all legal texts, references, metadata, and conclusions against the official Fedlex source before relying on them.
• Use at your own risk. The author assumes no responsibility or liability for errors, omissions, downtime, misinterpretation of legal materials, or any direct or indirect consequences arising from use of this project.
• Third-party source dependency. Availability and correctness depend in part on upstream Fedlex services, data formats, and network conditions, which may change without notice.

By using this project, you accept that it is a best-effort tool and not an authoritative legal system of record.