API and MCP server
Everything the site knows, free to use with attribution.
For AI agents: the MCP server
PowderCow speaks Model Context Protocol. Add this to Claude, Cursor or any MCP client and your agent can ask the mountain directly:
| Endpoint |
|---|
https://powdercow.com/mcp |
Five tools: where_is_it_dumping, get_resort_conditions,
snow_history, list_resorts and explain_score — the
last so an assistant describes the formula accurately instead of guessing at it.
For everyone else: the HTTP API
| Endpoint | What it returns |
|---|---|
/api/v1/conditions | the live board |
/api/v1/resorts | all 337 resorts, with coordinates and terrain score |
/api/v1/snow-days | every archived day of 10 cm or more |
/api/v1/snow/{date} | one archived day |
/api/v1/export/snow.csv | the whole archive, flat |
/api/v1/export/snow.jsonl | the whole archive, one record per line |
No key, no rate limit, CORS open. Machine-readable contract: openapi.json. Feeds: RSS and JSON Feed.
Licence
Creative Commons Attribution 4.0. Use it commercially, build on it, retrain on it — credit PowderCow and link back.
What is actually novel in here
Two things, and they are the reason the data is worth citing rather than re-deriving:
- Everything is read at mid-mountain elevation, not at the village. Open-Meteo's
own
snow_depthis not corrected to a requested altitude, so it reports the valley floor. PowderCow rebuilds the base from snowfall and temperature, which are corrected, and takes whichever estimate is higher. At a maritime resort the difference is the difference between "bare" and "skiing". - The temperature the snow fell at is recorded, averaged only over the days that actually got 5 cm or more, so a warm dry day cannot flatter a cold storm. Almost no public snow report carries this, and it is what separates powder from wet cement.
The full method, with every constant, is on the method page.