Patent Agent is where Patent & R&D Operations' AI actually runs — jurisdiction-aware filing forms for TR/EPO/US/WIPO, and ten dedicated reasoning endpoints for collision risk, prior art, claim charts, valuation, and more.
Claim structure and required abstract fields differ by patent office. Select a country and the form itself changes — field names, filing fees in the right currency, and the statute that governs it.
Every AI-calling screen in Patent & R&D Operations has its own dedicated endpoint — collision risk, prior art novelty, claim charts, FTO clearance, literature synthesis, grant compliance, license audits, and IP valuation. Each one's prompt and output schema stay independent, so a change to one screen's reasoning can never silently affect another's.
Nearly every screen runs the same two-stage flow. Stage 1 narrows a target spec, claim, or query down to its closest matches using local embeddings — that's Knowledge Engine, always on your own machine. Stage 2 is Patent Agent reasoning over that shortlist.
Every one of Patent Agent's ten endpoints applies the same routing logic once a request reaches the server — which only happens when you've chosen API or Local CLI in Settings → Agentic. Ollama, local-LLM, and web-LLM never send a request here at all.