Free for evaluation. Paid plans add higher call volume and the statistical surface area that traders, utilities, and researchers actually run their decisions on. Every plan shares the same calibrated response shape, the same per-day provenance, and the same verification receipts.
Plans
Credits debit per call, weighted by endpoint cost. Plans renew monthly; unused credits do not roll over.
Basic
$29 /mo
Pro
$99 /mo
Team
$499 /mo
Enterprise
Custom
Compare features
The parts that make a forecast trustworthy are universal. The calibrated intervals, the per-day provenance and the verification receipts are identical on the free tier and on Enterprise — we would rather you evaluate the real thing than a crippled version of it. What the paid tiers add is call volume, more keys, and statistical surface area: higher moments from Basic, window-comparison tests and mixture fits from Pro.
| Feature | Free | Basic | Pro | Team | Enterprise |
|---|---|---|---|---|---|
| Five core endpointsclimatology · forecast/point · seasonal · history/observations · forecast-vs-actual | ✓ | ✓ | ✓ | ✓ | ✓ |
| Confidence intervals (μ, σ, p05–p95) | ✓ | ✓ | ✓ | ✓ | ✓ |
| Per-day provenancesource, model cycle, station and network on every value | ✓ | ✓ | ✓ | ✓ | ✓ |
| Forecast verification receiptsin_p80 / in_p90 per row, plus rolling coverage and calibration error | ✓ | ✓ | ✓ | ✓ | ✓ |
| Rolling moments — mean, σ, min, maxon the trailing 30-day window of any history call | ✓ | ✓ | ✓ | ✓ | ✓ |
| Rolling higher momentsskew, kurtosis, p25/p75, IQR | — | ✓ | ✓ | ✓ | ✓ |
| Rolling-window overrideany span from 1 to 3650 days; free is served at the 30-day default | — | ✓ | ✓ | ✓ | ✓ |
| /v1/window/compareKolmogorov–Smirnov, Mann–Whitney, one-way ANOVA | — | — | ✓ | ✓ | ✓ |
| Granger causality between stationsdeseasonalized, AIC lag selection, ships its own measured false-positive rate | — | — | ✓ | ✓ | ✓ |
| Gaussian mixture fit + parameter return1–3 components by AIC, with KS goodness-of-fit | — | — | ✓ | ✓ | ✓ |
| Monthly creditsdebited per call, weighted by endpoint cost | 1,000 | 10,000 | 50,000 | 300,000 | Custom |
| API keysmint and revoke from your account at any time | 1 | 1 | 3 | 25 | 100+ |
| Support | Community | Priority email | Shared Slack | Direct engineering | |
| Custom priors & hierarchical modelsscoped per engagement | — | — | — | — | ✓ |
| SLA + dedicated WFO-region routing | — | — | — | — | ✓ |
Who picks which plan
Plan picks aren't really about credits — they're about which statistical surface you need a response to carry. Here is how the most common shapes of customer map onto the tier ladder.
You're sizing heating-degree-day positions and pricing weather options against a full distribution, not a point estimate. Integrate the prediction interval over your bracket and you have a probability to compare against the quoted price. Before you size anything on it, pull /v1/forecast-vs-actual for the station and window you care about and check our empirical coverage yourself — the interval is only worth trading if it holds up out of sample. On Pro, /v1/window/compare answers the follow-up question — is this June's tail the same population as last year's? — and a mixture fit tells you whether the window is one regime or two.
Recommended ProLoad planning, maintenance windows, and storm-response prep all want the same thing: a stress band you can plan against. Take p05 and p95 straight off the prediction interval for the daily band, and /v1/seasonal for the 1–12 month view. On Pro, /v1/window/compare settles "does this winter look like 2014?" with a distribution test rather than an eyeball. The 25-key Team allotment fits multi-region ops desks that want one key per site.
Recommended TeamFrost risk and irrigation calls depend on the worst-case low more than the mean. Ask for kind: "low" and plan against p05 rather than the point estimate. Rolling kurtosis and IQR on the history endpoint, from Basic up, give you the tail shape behind that number. /v1/climatology tells you what is normal for that day of year at that station, so you can see whether a cold night is genuinely unusual or just a normal spring.
Reproducibility is the whole problem. Every value carries per-day provenance — which source produced it, which model cycle, which station — so a result you publish can be traced back to its input years later. /v1/history/observations returns the same quality-flagged ASOS record we calibrate against, and /v1/forecast-vs-actual gives you our error series directly, so our calibration claim is something you can referee rather than take on trust.
Recommended TeamMost consumer UIs only need /v1/forecast/point and the confidence interval to render "78°F, give or take 5°" instead of a bare number. Per-day provenance gives your support team a clean answer when a user asks where the value came from.
Recommended BasicPick a sign-in email and password. You'll get your first key on the next screen, ready to call any of the five endpoints right away. Upgrade to a paid plan from any pricing card above.
Your API key (shown once — copy it now):
Send it as the X-Oasis-Weather-Key header on requests to
.