Searching patents by keyword is noisy: one concept, many phrasings. Classification assigns each patent to a technology code, so filtering by class captures the whole category, not a phrasing.
For monitoring, a class filter answers 'what's new in this technology area?' precisely.
USPC (U.S. Patent Classification) is the legacy system the USPTO used for decades.
CPC (Cooperative Patent Classification) is the modern joint US/European scheme and is far more granular.
Most current monitoring and search happens in CPC, which aligns US and international classifications.
To watch, say, photovoltaic mounting, you'd filter on a CPC class like H02S 20/xx. New grants or publications falling under that class become your feed.
This is the difference between 'new solar patents' (vague) and 'new patents in CPC H02S 20/30+ tracking/mounting' (actionable).
A structured feed exposes the class field on each record, so an agent can filter, cluster, and report new activity per class automatically.
That is the core of competitive technical intel: not 'what did they patent' but 'what changed in classes I care about'.
Cooperative Patent Classification — the modern US-EPO joint scheme.
Historic data uses it; current work is predominantly CPC.
They go from broad sections (H) down to very narrow groups (H02S 20/32), enabling precise filtering.