Patent Sentinel Evidence Service runs an adversarial, element-by-element analysis of a patent against an accused product and returns a verdict — but the verdict is screening, not adjudicated infringement.
What it is
Patent Sentinel Evidence Service is a Kcalbin service built for one task: given a patent and an accused product, it argues the infringement question from both sides and returns a structured answer. A prosecution brief and a defense brief are built independently, then a neutral judge weighs them and returns a verdict — SUPPORTED, NOT SUPPORTED, or INCONCLUSIVE — along with the controlling reason.
The output is not just a label. It includes an element-by-element breakdown: the plaintiff's and defendant's positions on each claim element, and the defendant's strongest non-infringement or invalidity argument stated at full strength, not softened. Every source cited in the analysis can be opened and checked, and the output is not meant to lean on inflated confidence scores standing in for evidence.
The service works the same way regardless of which side you're on. An assertion-side team and a defense-side team get the same kind of analysis, run the same way. You bring the patent and the accused product; Kcalbin only runs third-party analyses on what a requester brings, and does not expose its own portfolio targets to anyone.
It's also built to be called by machines first. There's a published discovery card and a live agent-to-agent task endpoint that takes a patent number, a defendant product, and a defendant company (claim numbers are optional), authenticated with a bearer token issued at subscription. A separate health endpoint is open without a key, so a buyer's own agent can confirm the service is running before anyone subscribes. Calling the task endpoint without a key returns an HTTP 401, by design. Keys come with a subscription — tell Kcalbin the volume you expect and whether you're working assertion-side or defense-side, and pricing comes back for that shape of use.
What it does not do
This is the part worth reading slowly, because the product is explicit about it.
A verdict from Patent Sentinel Evidence Service is not adjudicated infringement, and it is not legal advice. The service states its own operating rule plainly: similarity is never proof of claim practice. That is not a disclaimer appended at the end — it is the premise the whole analysis runs on.
A verdict reflects the openable record at the time of the run. It is screening and evidence analysis, not a ruling, and nothing it returns should be treated as a substitute for a lawyer's judgment about a specific dispute.
And a negative or inconclusive result is not a failure state. The product says this directly: NOT SUPPORTED and INCONCLUSIVE are valid, common, and useful outcomes — especially for a defense team trying to understand where an assertion is weak. Don't mistake "the tool found a verdict" for "the tool found infringement." Often what you're paying for is a clear-eyed NOT SUPPORTED with the reasoning behind it.
The practical read
If you're an in-house counsel or a patent operations team trying to triage a stack of potential assertions, or trying to pressure-test one you've received, Patent Sentinel Evidence Service gives you a structured first pass: both sides argued independently, the weak points named rather than hidden, and every citation left open for you to check yourself. Treat the SUPPORTED, NOT SUPPORTED, or INCONCLUSIVE label, and the reasoning behind it, as an input to legal judgment rather than a replacement for it. The service itself draws that line: it is screening and evidence analysis, not adjudicated infringement, and not legal advice.