Built for software teams. Screen your design against software patent claims in minutes, then engineer your way out. Three tools, one workflow.
Run a checkFind what is out there, check how close it is, engineer around it. Each tool feeds the next.
What it is: paste a patent (or just its number) and describe your product. You get a chart scoring every claim element against your design, and a distance score from 0 to 100.
What it is: after a check, the brainstormer proposes engineering alternatives built to break the closest claim elements. Each alternative is scored by the same engine, so you know it actually escapes.
What it is: describe what you are building and search every claim element we have captured. See the closest patents, then screen the top 8 in one batch pass.
Works on any granted US patent, software or not. Claims are claims: the parser and scoring engine do not care what field the invention is in.
Works on any domain, but the deepest vocabulary and the casebook precedents are software: data processing, networking, graphics, business methods, AI.
Software categories only today (G06F, G06N, G06Q, G06T, H04L, H04W). Other categories arrive as the corpus grows; the software corpus is where our data is deepest.
Input
Patent: US 12,726,443 B1, network traffic rate limiting (fetched automatically from the number)
Your design: Our API gateway checks request rates in a distributed key-value cache using a fixed window counter per user, and sheds burst traffic by returning 429s. It does not generate or refresh tokens.
Output
Likely clear (low confidence, review) Design distance 80/100
| Claim 1 · token refresh interval determination | absent |
| Claim 1 · dynamic adjustment on burst detection | different approach |
| Claim 2 · windowed counter per user | similar |
Real run against a live USPTO claim set. The similar limitation is why this stays in review instead of a flat all-clear.
Input
The chart’s closest limitation, fed in automatically: dynamic adjustment of the rate limit upon burst detection
Output
One brainstorm run returns up to 6 options, every one re-scored against the verbatim claim element before you see it:
| Token bucket, fixed refill · drains per request, refills on a timer, so the limit never adjusts to traffic shape | absent verified |
| Static tiered caps · three fixed rate tiers chosen at deploy time, no runtime adjustment at all | absent verified |
| Client-side shedding · clients self-throttle from response headers, server never detects a burst | different approach |
Each option comes with the strongest argument that it STILL infringes (a prosecutor attack) and manufacturing or implementation notes. Options that fail the engine check are flagged, not hidden.
Illustrative example. Options are split into proven patterns (built on verified design-arounds from past runs) and fresh ideas (divergent exploration).
Input
Description: a collapsible fluid bag for medical infusion, a multi-layer pouch that collapses as fluid empties, with a self-sealing fill port
Output
| US 12,602,087 B2 | 0.593 |
| US 12,608,084 B2 | 0.546 |
| US 12,737,713 B2 | 0.546 |
Then one click: Screen the top 8 against your design runs a full clearance check on each and ranks them by verdict.
Real search over the live corpus (2.5M+ claim elements). Search finds candidates; the batch screen scores them. Search cannot promise it finds everything.
A patent carries numbered claims, up to 20 of them, and each claim is built from limitations, the individual elements that legally define the invention. Infringing means practicing every limitation of a claim. Free plans screen the first 3 claims, including claim 1, the broadest. Paid plans screen all 20, where the narrow traps usually live.
| Free | Engineer $29/mo | Consultant $79/mo | |
|---|---|---|---|
| Clearance checks | 3/day | 100/day | 500/day |
| Claims screened per patent | first 3 | all 20 | all 20 |
| Design-around brainstorms | none | 20/mo | 100/mo, priority |
| Batch screen (top 8 patents) | not included | yes | yes |
| PDF export | not included | yes | branded + API access |
Quotas reset daily (checks) and monthly (brainstorms), US Eastern. Batch screening charges one check per patent screened.
A non-infringement opinion from a law firm runs about $14,000 per patent. A freedom-to-operate analysis runs $10,000 to $50,000. One third of venture-backed startups receive a patent demand letter. Engineers deserve a first answer before the lawyers bill the first hour.