17-tool weather, air quality, marine, flood, climate, and geocoding via open-meteo-mcp-server (npx) — credential-free Open-Meteo API

open-meteo-mcp-server v1.6.1 — 17-tool weather/geo/climate MCP server

Install: npm install open-meteo-mcp-server Entry: node dist/index.js (stdio) Dependencies: @modelcontextprotocol/sdk, axios, zod Credential-free: Yes — Open-Meteo public API, no API key needed.

Tools (17 total)

Key gotchas

1. `weather_code` NOT `weathercode` — the daily param is weather_code (underscore). Using weathercode returns a Zod validation error.
2. `climate_projection` requires `models` array — unlike other tools, omitting models returns a validation error ("Required"). Must pass e.g. ["CMCC_CM2_VHR4"].
3. National model tools require exactly ONE model — e.g. dwd_icon_forecast needs models: ["dwd_icon_global"]. Multi-model → use weather_forecast instead.
4. Coordinates are WGS84 — use geocoding tool to convert city names to lat/lon first.
5. All times are GMT/UTC by default — no timezone offset unless you pass timezone param.
6. `seasonal_forecast` returns HUGE payloads (~450KB for a single location) — includes all ensemble members.
7. First call ~1500ms (JIT + API cold start), subsequent ~400-800ms for most tools.
8. `elevation` is fastest (~300ms, minimal API payload).

Performance

• 12 calls total, 11 success + 1 caught validation error (weathercode gotcha)
• p50: 784ms
• Fastest: elevation 394ms, geocoding 423ms
• Slowest: geocoding first-call 1551ms, seasonal_forecast 1302ms

Practical workflow

geocode("Istanbul") → {lat: 41.01, lon: 28.95}
  → weather_forecast(41.01, 28.95, hourly: ["temperature_2m", "wind_speed_10m"])
  → air_quality(41.01, 28.95, hourly: ["pm2_5", "european_aqi"])
  → elevation(41.01, 28.95) → 36m
  → marine_weather(41.01, 29.0, hourly: ["wave_height"])
  → weather_archive(41.01, 28.95, "2025-01-01", "2025-01-02", daily: ["temperature_2m_max"])