Patent-Alerts is an MCP endpoint that turns USPTO publication data into clean, dated, machine-readable records. Search by number, assignee, or technology class — get JSON your pipeline can act on.
search_patentLook up a patent document by number, assignee, scientist, or USPTO classification class.
new_grantsReturn recently granted U.S. patents as a dated, machine-readable signal feed.
monitor_classWatch a USPTO technology class (class/cpc number) and receive new activity in it.
assignee_lookupList patents grouped by assignee — competitors, portfolio scoping.
{
"query": "solar panel",
"count": 2,
"results": [
{
"patent_number": "US12,345,678",
"title": "Method for photovoltaic panel efficiency monitoring",
"grant_date": "2026-09-07",
"assignee": [
"Solar Dynamics Inc."
],
"cpcs": [
"H02S 50/10"
]
},
{
"patent_number": "US12,345,600",
"title": "Dual-axis solar tracking system",
"grant_date": "2026-08-24",
"assignee": [
"HelioTrack LLC"
],
"cpcs": [
"H02S 20/32"
]
}
],
"note": "Illustrative shape of the tool response \u2014 values shown as examples, not a live query result."
}
Illustrative shape of the tool response — example values, not a live query. Live calls hit the published USPTO open dataset and surface real records.
Tools return structured JSON (see examples below): patent numbers, title, dates, assignee, and classification. You do not parse web pages; you get clean signals an agent can act on.
U.S. Patent & Trademark Office open datasets. Like all public-source tools, availability of the upstream endpoint can vary; the tool reports errors honestly rather than fabricating records.
Yes — monitor_class lets a workflow subscribe to a USPTO class and detect new activity, the kind of signal used for competitive IP intelligence and R&D pipeline feeds.
Both. The landing page is for humans; the /mcp endpoint, llms.txt and /.well-known/agent.json are the machine-readable surface agents consume directly.