Use Lex with your AI

your model + our evidence, MCP endpoint, one line to connect

Lex is MCP-native: you bring your AI, Lex brings the evidence. Your model asks the 10 Lex tools for the law as it stood on a date, and composes its answer from returned text, dates and hashes, Lex itself never interprets anything.

Connect in one line

Claude Code

claude mcp add --transport http lex https://law.soufien.lu/mcp

Claude Desktop / any MCP client, add a remote MCP server:

{ "mcpServers": { "lex": { "url": "https://law.soufien.lu/mcp" } } }

What a conversation looks like

    You     What did Luxembourg data-protection law require about breach
            notification in March 2019?

    AI  →   search("notification violation données", as_of: 2019-03-15)
        ←   loi du 1er août 2018 (lex_id lu-legilux:loi-2018-08-01-a686…), recueil protection des données
    AI  →   as_of("lu-legilux:recueil-protection_donnees", "2019-03-15")
        ←   the exact text valid that day + interval + sha256 + signed stamp

    AI      "As of 15 March 2019, the applicable framework was … [quotes the
            retrieved text; cites validity 2018-08-20 → 2019-… ; links provenance]"

The 10 tools

as_of · timeline · in_force_on · diff · search · article_history · provenance · coverage · cited_by · changes_in_period , read-only, deterministic, every response carries its dates, its hash, and an honest refusal (no_version_for_date, text_withheld) when Lex cannot know.

Azure AI Foundry agents

Foundry's Agent Service speaks remote MCP natively, point an agent at this endpoint and it gets all 10 tools (no key needed; leave approvals on for writes-free comfort):

{ "type": "mcp", "server_label": "lex", "server_url": "https://law.soufien.lu/mcp", "require_approval": "never" }

Want to try it without installing anything? The capped built-in playground runs the same tools. Prefer no AI at all? Everything is also a permalink.