GribStream Blog
Public catalog endpoints now available
GribStream now exposes public catalog endpoints to list datasets, inspect parameter metadata, and resolve shared cross-dataset selectors.
On April 19, 2026, we made the GribStream catalog public.
Up to now, the public API has mostly been about getting data from /timeseries and /runs. We also had a few metadata helpers built mainly to support dashboards, but they were not something we wanted customers building against long term.
This release changes that.
What is now available
GET /api/v2/catalog/datasetslists datasets together with public metadata such as provider, cadence, archive coverage, tags, source links, and model-page links.GET /api/v2/catalog/datasets/{dataset}/parameterslists the parameters available for one dataset.GET /api/v2/catalog/datasets/{dataset}/parameters/{parameter}returns the full detail for one exact parameter code, including selector variations.GET /api/v2/catalog/shared-parametersandGET /api/v2/catalog/shared-parameters/{parameter}expose a curated cross-dataset layer for applications that want one logical signal mapped onto multiple models.
One detail worth calling out: parameter short names are returned in their exact upstream casing. If a dataset exposes 100u, the catalog returns 100u, and that is the value to reuse in forecast queries.
Who should use it
If you already know you want something like GFS TMP at 2 m above ground, nothing changes. The simplest path is still to call the data endpoint directly. The Quickstart continues to start there.
The catalog is more useful when you are building:
- a dataset or parameter picker
- a multi-model dashboard
- internal tooling
- generated clients or SDK helpers
- an application that needs to discover selectors instead of hardcoding them
I added an advanced catalog section to the Quickstart for that workflow, after the regular data-query examples.
Why we wanted this in place
We are about to make these endpoints part of the stable public surface, so it was important to get a proper foundation in place now rather than keep extending the earlier ad hoc metadata helpers.
This should make it much easier to build on top of GribStream, whether you are wiring up your own application, exploring datasets like NBM or IFS Operational, or generating integrations from the OpenAPI spec.
It also gives us a cleaner base for the next step: using this catalog to build MCP and skill documentation so AI tools can discover how to query GribStream correctly instead of relying on trial and error.