Open-Meteo API Skill

An Agent Skill that teaches AI agents (Claude Code, Claude.ai, and other skill-compatible agents) how to use the free Open-Meteo weather API.

Open-Meteo blends national weather models (NOAA GFS, DWD ICON, ECMWF, Météo-France, JMA...) into a single API — no API key, no signup for non-commercial use, with forecasts up to 16 days and historical data back to 1940.

What the skill enables

Once loaded, an agent can reliably:

• Answer "weather in X" in one step with the bundled get_weather.py script (zero dependencies, Python stdlib only)
• Resolve a place name to coordinates with the Geocoding API
• Fetch current conditions (temperature, feels-like, humidity, wind, weather code)
• Get hourly and daily forecasts up to 16 days (rain probability, UV, sunrise/sunset...)
• Interpret WMO weather codes into human-readable conditions and icons
• Pull historical weather from 1940 onward via the Archive API
• Check air quality (PM2.5, US/European AQI, pollen)
• Query marine (waves, sea temperature), elevation, flood, and climate projection APIs
• Avoid real-world pitfalls (GMT default timezone breaking daily aggregation, missing results key on failed geocoding, parallel-array response format, archive's ~5-day delay)

All request/response examples in the skill were verified against the live API.

Repository layout

skills/
└── open-meteo-api/
    ├── SKILL.md                     # Entry point: workflow, forecast + geocoding, WMO codes, tips
    ├── scripts/
    │   └── get_weather.py           # Zero-dependency CLI: place name → formatted weather (stdlib only)
    └── references/
        ├── weather-variables.md     # Full current/hourly/daily variable catalog + WMO code table
        └── other-apis.md            # Archive, air quality, marine, elevation, flood, climate, ensemble

Installation

Claude Code (plugin marketplace — recommended)

This plugin is distributed through the NanookAI/skills marketplace. Inside Claude Code, run:

/plugin marketplace add NanookAI/skills
/plugin install open-meteo-api@nanookai-skills

The skill is then available in every project, and /plugin marketplace update picks up new versions.

Claude Code (per-project)

Copy the skill folder into your project's .claude/skills/ directory:

cp -r skills/open-meteo-api /path/to/your-project/.claude/skills/

Claude Code (personal, all projects)

cp -r skills/open-meteo-api ~/.claude/skills/

Claude.ai / other agents

Package the skill folder (respecting .skillignore) into a .skill bundle, or upload the skills/open-meteo-api/ folder contents wherever your agent platform accepts skills. The skill is self-contained; the bundled script needs only a Python 3.8+ standard library (no packages to install), and everything else is plain documentation.

Try it

After installing, ask your agent things like:

• "What's the weather in Taipei right now?"
• "Will it rain in Tokyo this weekend?"
• "Compare this July's temperatures in Kaohsiung with last year's"
• "What's the UV index and air quality in Taichung today?"
• "When is sunset in Reykjavik tomorrow?"

About the API

• Forecast endpoint: https://api.open-meteo.com/v1/forecast
• Free for non-commercial use (fair-use limit ~10,000 requests/day); commercial use requires a paid API plan — see open-meteo.com/en/pricing
• Open source (AGPLv3) and self-hostable — see github.com/open-meteo/open-meteo
• Weather data licensed under CC BY 4.0 (attribution required)

License

The skill documentation in this repository is provided as-is. Open-Meteo itself is an independent open-source project — see open-meteo.com for its license and terms of use.