Endpoint
api.invoket.com/weather/forecast/exceedance
Before your agent commits to a rain-or-shine decision two to five weeks out — a parametric payout, a shipping window, a harvest or concrete slot: how many NOAA GEFS ensemble members cross your threshold on each day of the ensemble tail. Ask 'at least 30 mm of rain in a day', 'wind at or above 15 m/s' or 'a day at or below 0 deg C' for any GPS point or city, and get one counted fraction per UTC calendar day with the member count behind it. Ask days= (up to 20) for the whole tail. This is what the ensemble MEAN served by GET /weather/forecast cannot answer: averaging the members smooths extremes away, so a mean day essentially never crosses a threshold. What is served here is a raw counted fraction, not a calibrated probability — no bias correction or statistical post-processing is applied, and members_total is served on every day so the number can be audited.
Score
The score is built from observations we collect ourselves: how the endpoint responds, what the envelope contains, how stable the price and the recipient are, who pays and how often. It is deterministic — the same input always produces the same result.
We do not publish the exact methodology yet. If you own this endpoint, write to us and we will show you, item by item, what drags the score down.
Recent probes
| When | Code | Latency | Scheme and network | Recipient | Price | Findings |
|---|---|---|---|---|---|---|
| 3h ago | 402 | 401 ms | exact eip155:8453 |
0x4465…28ee | 0.010000 | clean |
| 4d ago | 402 | 420 ms | exact eip155:8453 |
0x4465…28ee | 0.010000 | clean |
| 4d ago | 402 | 413 ms | exact eip155:8453 |
0x4465…28ee | 0.010000 | clean |
Reference
| URL | https://api.invoket.com/weather/forecast/exceedance |
| Sources | well-known |
| Recipient per catalogue | 0x4465825B0ac713aA2Fa07Ea29023020804A128ee |
| Machine access | /v1/endpoint · /full · badge |
Is this your endpoint?
The score is built only from public observations. If something we recorded is wrong — say our prober hit you during maintenance — write to us. We will recheck and publish your reply next to the score.
We do not delete accurate observations. The prober makes at most one request per second per host and only asks for the payment challenge — it never pays and never takes content.