Return the current credit balance and subscription status for the authenticated AI Patent Search Generator account. Use this before calling paid tools (dossier, oa_analyze, query, search) to verify credits are available.

Fetch a comprehensive intelligence dossier for a patent by publication number. Returns bibliographic data (title, assignees, inventors, dates, legal status), independent claims, backward + forward citations, patent family, CPC classifications, examiner info, and similar documents. Costs 3 credits on a fresh fetch; subsequent calls within 24 hours read from cache and are free. The headline tool — most patent-analysis workflows start here.

Retrieve the USPTO file-wrapper documents (office actions, responses, amendments, IDS, notices) for a US patent or application. Free. Returns each document's ID, mail date, category, and short description — IDs can be passed to oa_analyze for AI analysis.

AI-analyze a USPTO Office Action document to extract rejection grounds (102, 103, 112, etc.), cited prior art, and suggested response arguments. Two invocation forms: (1) pass only patentNumber and the server auto-picks the most recent Office Action for the patent; (2) pass both applicationNumber and documentId (from prosecution output) for explicit selection. First 5 analyses per application are free; subsequent analyses cost 1 credit each.

Get the assigned examiner's name, art unit, total applications handled, allowance rate, and average pendency for a US patent. Free. Useful for prosecution strategy and risk assessment.

Generate a single optimized Boolean search query for Google Patents from a natural-language description of an invention or technology. Returns the Boolean string only — does NOT execute the search. Use search if you want ranked hits instead. Costs 1 credit.

Execute a patent search against Google Patents server-side and return ranked hits. Given a natural-language description and a search strategy, the server: (1) extracts technical concepts, (2) generates Boolean queries per the chosen strategy, (3) runs each query against Google Patents, (4) dedupes results by publication number. Strategies: 'telescoping' (3 queries: broad/moderate/narrow — recommended default), 'onion-ring' (layered queries adding one concept at a time), 'faceted' (multiple two-concept pair queries). Costs 1 credit.

Return Google Patents' 'similar documents' ranking for a given patent — a list of patents that Google considers technically related based on its own semantic models. Free. Distinct from search (which generates new queries from a description); this tool finds patents related to an existing one.

Return backward and/or forward citations for a patent. Backward = patents this one cites (its prior art). Forward = patents that cite this one (downstream impact). Each citation includes whether it was examiner-cited (load-bearing on patentability) vs applicant-cited. Free. Lighter payload than dossier when you only need citation lists.

Return the patent family for a given patent — continuations, divisionals, and national counterparts (foreign equivalents). Useful for confirming worldwide IP coverage and identifying related applications still pending. Free.

Look up a Cooperative Patent Classification (CPC) code and return its section, class, subclass, and group context. Accepts any level of code — section letter (e.g. 'H'), class ('H01'), subclass ('H01M'), main group ('H01M10/00'), or subgroup ('H01M10/0525'). Free. v1.0 dataset covers all sections + ~80 common subclasses; subgroup-level descriptions land in v1.1.