Answers about PatentAgility’s USPTO examiner analytics, U.S. patent family review tools, claim analysis workflows, support search, accounts, data handling, and system limits.
PatentAgility combines USPTO public data with local NLP, AI, and ML workflows to automate tasks that are otherwise tedious, repetitive, or time-consuming in patent practice. Except for retrieving public USPTO-related material, the substantive processing is local to our own systems: this website is not just a wrapper around outside online LLMs. The code is open source and updated regularly.
Examiner Analytics. This tool turns large volumes of USPTO-related examiner, art unit, group, and tech-center data into a browsable analytics layer. It aggregates roughly 12.7 million office-action records tied to millions of prosecution events, then computes descriptive metrics such as observed grant ratios, office-action counts, and rejection trends so you can move from raw public data to a much faster strategic read on examination patterns.
Claim Amendment History across U.S. Family. This tool starts from a patent or application number, resolves the related U.S. family, retrieves prosecution-related claim history across those U.S. family members, and compares independent claim language over time. In practice, it is doing a substantial data-linking and text-comparison task across a connected U.S. family layer of more than 12.7 million applications and millions of family relationships that would otherwise take a great deal of manual review.
Compare Claims Across U.S. Patent Family. This tool identifies granted U.S. family members, collects the latest granted independent claims from each, aligns them side by side, and uses local AI summarization to describe the claim language and the major differences across the U.S. family. It turns a dense, multi-document comparison problem into a faster portfolio-level view of scope movement, drawing on a grant corpus of roughly 6.6 million issued patents.
Find Unclaimed Subject Matter in U.S. Family. This tool resolves the related U.S. family, loads the seed specification, extracts concepts from both the specification and the U.S. family claims, and then scores whether those specification concepts appear to be covered in the claims. It combines large-text processing with NLP-driven concept extraction on top of a patent-text layer that spans more than 100 million claim-text rows, helping surface potentially weakly claimed or unclaimed material that is easy to miss in manual review.
Find Support in Specification. This tool loads specification text for the identified patent or application, breaks it into searchable units, and runs a hybrid retrieval pipeline against your query using both lexical and semantic search signals before reranking the candidates. In other words, it uses modern NLP retrieval methods to search a long technical disclosure more intelligently than a simple keyword scan, with the broader platform built on more than 6 million detailed patent descriptions.
Antecedent Basis Checking. This tool parses claim text, tracks introduced terms and later references to them, and flags likely antecedent-basis or referential issues using rule-based NLP heuristics. It is designed to automate the kind of detail-oriented linguistic review that patent professionals often do manually, line by line.
Linguistic Claim Analysis. This tool parses claim text with syntactic and rule-based segmentation methods, identifies relationships among limitations, and renders a structured claim chart. It uses NLP to convert dense claim language into a more inspectable structure, which can make long claims much easier to analyze, explain, and revise.
Claim Summarizer. This tool parses a submitted claim set, groups claims by status and dependency signals, identifies the independent pending claims, and then uses local AI summarization to generate short plain-English summaries of those independent claims. It is intended to compress a complicated claim set into a faster high-level read without losing the underlying claim structure.
Some of the heavier tools use expensive models and larger processing pipelines. Accounts let us rate-limit abuse, reduce spam, and stop repeated or automated misuse from dragging down the system.
They also let us tie jobs to users for diagnostics, performance monitoring, and basic usage tracking.
This system has real limits. A short version:
USPTO data can be messy, incomplete, or misleading. OCR can be bad, typos can survive, dates can be wrong, and malformed XML can create bad associations. We cache that data locally for speed, but we do not fully clean it, so bad source data can become bad output. Older OCRed claims are especially prone to header noise and obvious OCR artifacts.
We rely on local AI models. That keeps substantive processing under our control, but it also means weaker output than top commercial LLMs. Plain-English summaries can be awkward or wrong.
The code can have bugs. PatentAgility is open source and updated often, but that does not make it error-free. Do not use it as your only source of authoritative fact.
In addition to ordinary web server logs (e.g., IP addresses and timestamps), if you create an account we store your e-mail address, a hashed and salted password, your job title, your company, your organization type, and basic account timestamps such as account creation and last login. We also store basic job metadata tied to your account, such as the job ID, job category, timestamps, stage, elapsed time, and whether the job completed successfully.
We do not store the substantive text you submit for a job, such as claim text or support-search queries. One practical exception applies: if we retrieve public USPTO documents and temporarily cache them to avoid repeated downloads or OCR work, those public-source materials may be stored internally for a period of time.
As a general matter, none. The substantive model and analysis work is performed on systems we control, and we do not send your claim text, search text, or similar substantive inputs to outside AI providers. In some workflows, however, our systems may retrieve public USPTO materials or query USPTO-related public data sources needed to complete the task.
Yes. PatentAgility was designed and coded by one of KellDann's founding partners, Kirk Sigmon. Part of the reason this tool exists is to show that KellDann's attorneys are genuinely technically capable in AI/ML and software. There was no outsourcing of the core system.
Tell Kirk. PatentAgility is experimental, and bug reports or examples of incorrect output are useful.