Claim History Tool
Patent Family Timeline & Claim Evolution
See how independent claims moved from filing to grant across a family, with AI-generated narrative framing around prosecution turns.
AI, LLM, and NLP Tools for Patent Analysis
AI patent analysis tools, coded by lawyers.
PatentAgility combines large-scale USPTO data, secure and local language models, dense retrieval, reranking, and Natural Language Processing ("NLP") techniques to compress hours of prosecution review into minutes.
6.6M
granted patents indexed
12.7M
applications connected
100% Local
no outside LLMs, no input storage
WARNING:
These beta tools are limited by data availability, OCR quality, retrieval noise, and model limitations. They are built to accelerate expert review, not replace review by a patent lawyer.
Tools
Claim Strategy Tool
Compare Claims Across Patent Family
Line up granted independent claims side by side and let the system summarize how scope shifts from one family member to the next.
Support Search Tool
Find Support in Specification
Use hybrid retrieval and reranking to surface the paragraphs most likely to support a claim concept, not just exact keyword matches.
Drafting QA Tool
Antecedent Basis Checking
Scan claims for likely referential breakdowns and drafting inconsistencies before they trigger examiner friction or internal churn.
Claim Language Tool
Linguistic Claim Analysis
Break down difficult claims into diagram-style structures so the operative action, object, dependencies, and alternatives become much easier to review.
Claim Summary Tool
Summarize Claim Set
Parse a pasted claim set, count amendments and cancellations, classify the pending claims, and generate plain-English summaries for the independent claims.
You can't beat free and open-source.
Price profile
Uses external LLMs?
Open source visibility
Patent Attorney Involvement
Data privacy
Data processing approach
Speed
PatentAgility
Free.
No. Everything is local, and on purpose.
Yes - check it for yourself. The only stuff we don't surface relates to UI, our API keys, etc.
Yes - all systems developed by patent barred lawyers based on their real workflows.
Strong: we don't store anything other than public USPTO data and server logs.
Best-in-class, informed by academic research. No LLM "magic."
1-2 mins per run.
Other Patent AI Tools
$$ - Thousands a year, if not more.
Extensive - often reliant on ChatGPT, Gemini, or other tools outside their control.
Not at all - you never have insight into what is acutally happening.
Rarely - often "designed" by non-lawyers through imprecise and unreliable LLM prompts.
Uncertain - they promise security, then fire your data off to another vendor via an API.
"Let's prompt this LLM until it does what we want it to do."
Varies depending on tool.
Manual Attorney Review
$$$, especially at large/overbloated law firms.
Probably, but they might not admit it.
Not unless you buy them a drink first.
Yes, but not guaranteed - many firms rely on non-patent barred associates and paralegals.
Totally dependent on their IT stack.
Human mental processes and Microsoft Office.
Often slow - they're billing hours, after all.
FAQ
How does this system work?
PatentAgility uses a combination of the USPTO's open data systems, in combination with cutting-edge (but local and secure) NLP, AI, and ML techniques, to perform tasks that are normally very time-consuming and difficult. Critically, except for pulling public data from the USPTO, all processes are local: that is, this system is not a wrapper for online LLMs. Also, for your own validation and comfort, the code is open-source and maintained regularly.
What are the main limitations of this system?
A non-exhaustive list:
This system is reliant on USPTO data, which has limitations. The USPTO's open data portal is absolutely fantastic, but patent data is messy - OCR processes might have been imprecise, typos might be preserved, etc. This system uses a large local cache of that information for speed, but no attempts have been made to "clean" the data. For example, older OCRed patent claims often contain content from headers and understandable OCR artifacts. Note also that the USPTO often rate-limits requests, so any retrieval steps on our end can be somewhat slow and unreliable.
This system relies on local AI models. They are not as powerful as commercially-available LLMs, such as ChatGPT or Gemini. That's on purpose - we control all data in and out - but it does come with an accuracy penalty. Thus, for example, plain English summaries of patent claim changes are sometimes a little linguistically awkward.
There's always the risk of code errors. The code for PatentAgility is open source and routinely updated. That said, there's always a very real risk that it has errors, mistakes, or the like. It should not be used as the only source of authoritative fact on any project.
Was this system really made by lawyers?
Yes, 100% - by one of KellDann's founding partners, Kirk Sigmon. Part of the reason this tool exists is to show how KellDann's attorneys are truly skilled in AI/ML. There was no outsourcing or other strategies used.
What if I find an error?
Tell Kirk!