Drop your API key
into Claude.

The thirty-second install.

Claude Desktop. Open ~/Library/Application Support/Claude/claude_desktop_config.json (or the equivalent on Windows / Linux) and add:

{
  "mcpServers": {
    "lexapi": {
      "command": "npx",
      "args": ["-y", "@lexapi/mcp"],
      "env": { "LEXAPI_API_KEY": "lex_..." }
    }
  }
}

Restart Claude, then ask "what tools do you have?" — you'll see the ten lex_* tools listed.

Claude Code. One command:

claude mcp add lexapi --env LEXAPI_API_KEY=lex_... -- npx -y @lexapi/mcp

Cursor or any other MCP client. Point it at npx -y @lexapi/mcp with the same env var. If your client speaks MCP, the install is identical.

What your model can now do.

Ten tools, grouped by what they're for:

Find the right document.

• lex_search — keyword search with structured filters (text scope, date, document type, author, language). Returns CELEXes, titles, dates.
• lex_recent_documents — the last week's Official Journal publications, by type and language.
• lex_semantic_case_law / lex_semantic_legislation — ask in natural language, get back the most relevant cases or articles.

Read it.

• lex_get_document — full parsed structure: title, metadata, individually-addressable articles, tables, annexes.
• lex_get_metadata — title, type, dates, ECLI/ELI, keywords — without the body.
• lex_get_document_by_url — paste a EUR-Lex URL, get the document.

Follow the citations.

• lex_cited_by — everything that cites this document, grouped by edge type (reference, amendment, repeal, implementation).
• lex_cites — the inverse: everything this document cites.
• lex_citation_network — both directions in one call, with per-edge-type counts.

A real query, end to end.

You have the MCP installed, and you ask Claude:

Summarise how the GDPR has been amended since it came into force. For each amendment, give me the CELEX and a one-sentence description.

Claude already knows the GDPR's CELEX (32016R0679). It calls lex_cited_by for the amendments, fetches each one's metadata, then sorts and summarises. A few seconds, a handful of tool calls — and a sourced answer with real CELEX numbers instead of a confident guess. The whole thing costs around 30 credits: about 6% of the Free monthly pool, 0.1% of Starter. Cheap enough to let the agent be exhaustive.

Why it beats pointing a model at EUR-Lex directly.

EU law is citation-heavy, multi-lingual, and served as HTML that shifts without notice. Ask Claude about a directive through LexAPI and it gets the parsed structure, not raw HTML; ask about amendments and it gets graph edges typed by what kind of amendment they are; ask in French and it can fetch the French text. The model stays focused on the legal question instead of fighting the source.

Pricing & feedback.

Semantic search and the citation graph are just tools with higher credit weights — no separate "AI features" plan. Every tier gets every tool; the pool size is what scales with price. Free includes 500 credits a month, enough for roughly one hundred semantic queries. The credit model itself is covered in the pricing post.

The package is @lexapi/mcp, source open on GitHub. If a tool returns something the model can't parse, or you want a tool that doesn't exist yet, email [email protected] with the details.