OpenMeteo

OpenMeteo

Typed Open-Meteo weather client for Deno, Bun, Node.js, browsers and Cloudflare Workers: the Forecast, Ensemble (every member), Previous Runs, Historical Forecast, Single Runs and Historical Weather (archive) APIs, with an explicit model choice (GFS, HRRR, ECMWF IFS, ICON, GEM, ...) and any hourly, 15-minutely, daily or current variable by name. A lightweight, validated alternative to the openmeteo SDK.

Overview

import { OpenMeteo } from '@tundraconnect/open-meteo';

const client = new OpenMeteo();
const [centralPark] = await client.forecast({
  latitude: 40.7794,
  longitude: -73.9692,
  hourly: ['temperature_2m'],
  minutely15: ['temperature_2m'],
  models: ['ncep_hrrr_conus', 'gfs_seamless', 'ecmwf_ifs025'],
  timeformat: 'unixtime',
});
console.log(centralPark?.hourly?.temperature_2m_ncep_hrrr_conus);

Method	API	What it answers
forecast()	Forecast /v1/forecast	The latest forecast, per model
ensemble()	Ensemble /v1/ensemble	Every ensemble member (31 GEFS, 51 ECMWF ENS, ...), not the mean
previousRuns()	Previous Runs	What the model forecast 1–7 days earlier for each hour
historicalForecast()	Historical Forecast	Archived first-hours forecasts, back to 2021/2022
singleRun()	Single Runs	One model run exactly as issued (run: '2026-11-05T00:00')
archive()	Historical Weather	Reanalysis (ERA5 and others) back to 1940

Nothing is resampled, merged or rounded. Every column Open-Meteo sends comes back as-is — every model, every ensemble member, every null — with the grid cell, elevation, timezone and units. Several latitude/ longitude values request several locations in one call; every method returns one entry per location, in request order. With several models, Open-Meteo suffixes each column with the model (temperature_2m_gfs_seamless); columns() splits names back into variable, ensemble member and model. Anything not modelled as an option goes in params under Open-Meteo's own name.

Free and paid hosts. Without auth the client uses the free hosts (api.open-meteo.com, ensemble-api.open-meteo.com, ...), which are for non-commercial use only and rate-limited (600 calls/minute, 5,000/hour, 10,000/day and 300,000/month per IP at the time of writing). Commercial use needs a paid plan: pass its key as auth and every call goes to the matching customer-* host with an apikey parameter. hosts overrides any host (a self-hosted Open-Meteo, a proxy).

Out of scope (for now)

• The geocoding, elevation, air-quality, marine, flood, climate and satellite-radiation APIs.
• The FlatBuffers (format=flatbuffers) and CSV/XLSX outputs; this client reads JSON.

Documentation

Topic	Description
API	Configuration, hosts, every method and query
Errors	Error codes and diagnostic metadata
Schemas	Public Guardian schemas and inferred types

Upstream

• Forecast API reference
• Ensemble, Previous Runs, Historical Forecast, Single Runs and Historical Weather
• Pricing and API keys

Installation

Deno:

deno add jsr:@tundraconnect/open-meteo

Bun:

bunx jsr add @tundraconnect/open-meteo

Node.js:

npx jsr add @tundraconnect/open-meteo

Quick Start

Record what three models and the GFS ensemble say about tomorrow's high in Central Park, and what GFS said about it yesterday:

import { columns, OpenMeteo } from '@tundraconnect/open-meteo';

const client = new OpenMeteo();
const station = { latitude: 40.7794, longitude: -73.9692 };

const [deterministic] = await client.forecast({
  ...station,
  hourly: ['temperature_2m'],
  models: ['ncep_hrrr_conus', 'gfs_seamless', 'ecmwf_ifs025'],
  temperatureUnit: 'fahrenheit',
  timezone: 'America/New_York',
  timeformat: 'unixtime',
});

const [ensemble] = await client.ensemble({
  ...station,
  hourly: ['temperature_2m'],
  models: ['gfs_seamless'],
  temperatureUnit: 'fahrenheit',
});
for (const c of columns(ensemble!.hourly!, ['temperature_2m'])) {
  console.log(c.model, c.member, Math.max(...c.values.map(Number)));
}

const [earlier] = await client.previousRuns({
  ...station,
  hourly: ['temperature_2m'],
  previousDays: [1],
  models: ['gfs_seamless'],
});
console.log(
  deterministic?.hourly?.time.length,
  earlier?.hourly?.temperature_2m_previous_day1,
);

License

MIT. Part of Tundra Connect, typed vendor API clients for Deno, Bun, Node.js and Cloudflare Workers.