Forecasts
Up to 9 days, hourly detail, ~6000 Swiss locations.
MeteoSwiss LLM Tools
Swiss weather data for AI assistants, powered by MeteoSwiss Open Government Data — the same data behind the MeteoSwiss app and website. Free, no API key.
It comes in two forms that answer the same questions: an MCP server with structured tools (hosted at meteoswiss-mcp.ars.is, or run your own) and an Agent Skill that needs no server at all.
Current weather
~300 stations, updated every 10 minutes.
Pollen
Seven species from ~15 monitoring stations.
Climate series
NBCN series going back decades, plus MeteoSwiss website search.
Pick your path
Section titled “Pick your path”Use itConnect Claude, Cursor or any MCP client. Tools, example prompts, data sources and limits.
Run itDocker or npm, every environment variable, caching, health and metrics, upgrades.
Build on itArchitecture, tests, the release process, and how to contribute.
See the evidenceEval runs and the skill-vs-MCP case study behind the design decisions.
Quick start
Section titled “Quick start”claude mcp add --transport http meteoswiss https://meteoswiss-mcp.ars.is/mcpThen ask: “Wie wird das Wetter in Bern morgen?” — other clients are covered in Connect to the hosted server.
Source and issues: github.com/eins78/meteoswiss-llm-tools · Software licence CC0-1.0 · Data: MeteoSwiss Open Government Data. This is a community project, not an official MeteoSwiss service.