I Published an MCP Server for Swiss Public Data. 68 Tools, 20 Modules, 867 Tests.

I shipped mcp-swiss — a Model Context Protocol server that gives any AI assistant direct access to Swiss open data. No API keys. No registration. Just npx mcp-swiss and you're connected to trains, weather, companies, parliament, earthquakes, and 15 other domains. It started at 37 tools across 9 modules. A few sprints later, it's 68 tools across 20 modules , backed by 867 unit tests . This is the story of building it, scaling it, and what I'd do differently. Wait — What's MCP? Model Context Protocol is an open standard that lets AI assistants call external tools. Think of it as a USB-C port for AI: one protocol, many data sources. Instead of pasting data into your prompt, the AI calls a tool directly: User: "Next train from Zürich to Bern?" AI → calls get_connections("Zürich HB", "Bern") AI: "The next train is at 14:32, arriving 14:58. Platform 31." The AI picks the right tool, calls it, and uses the result in its answer. No copy-paste, no hallucinated schedules. What mcp-swiss Covers T