LexDania MCP Server

A Model Context Protocol (MCP) server running on Cloudflare Workers. It enables AI models to query, structurally inspect, and perform RAG-based Q&A over Danish legislation published on Retsinformation.dk using their underlying LexDania XML.

Getting Started

Prerequisites

• Node.js & pnpm

• Cloudflare Wrangler CLI (pnpm install -g wrangler)

• Gemini API Key

Installation

pnpm install

Local Development

1. Create a .dev.vars file in the root:

   GEMINI_API_KEY=your-gemini-api-key
   FILE_SEARCH_STORE=fileSearchStores/your-store-id
   MCP_AUTH_TOKEN=your-secret-bearer-token

2. Start the development server:

   pnpm dev

Deployment

pnpm deploy

Related MCP server: MCP Server Legifrance

Connection

Add the following to your MCP client configuration (e.g., claude_desktop_config.json):

{
  "mcpServers": {
    "lexdania": {
      "url": "https://lexdania.pacmate.dk/mcp",
      "headers": {
        "Authorization": "Bearer <your-mcp-auth-token>"
      }
    }
  }
}

Features & Tools

Exposes the following tools to MCP clients:

• lexdania_ask_document: Free-text Q&A over Danish legislative documents using Gemini File Search.

• lexdania_query_document: Execute exact, namespace-aware XPath queries against a law's LexDania XML. Returns matched XML structures and aligned plain-text strings. The default (unprefixed) namespace is mapped to d:.

• lexdania_profile_document: Get a structural profile (root tag, schema variant, tag counts, depth, and namespaces) of a law's XML.

• lexdania_metadata: Fetch and parse JSON-LD metadata for a Danish legislation document by its ELI path or URL (returns title, document type, dates, ministry, status, and relationship edges).

• lexdania_compare_structure: Diff elements and attributes across multiple documents to identify parser drift and schema variants.

Project Structure

├── src/
│   ├── index.ts        # Worker entrypoint (routing & auth)
│   ├── http/           # HTTP authentication logic
│   ├── mcp/            # MCP server composition & tools
│   │   └── tools/      # Tool implementations (ask, query, structure, metadata, compare)
│   └── services/       # Core services (Gemini store, XML parsing)
├── test/               # Vitest suite
└── wrangler.jsonc      # Cloudflare Workers configuration

License

MIT