IronCrow AI - AI Tools for Patent Professionals

Check Antecedent Basis

Analyzes patent claims for antecedent basis issues. Returns JSON with a per-severity summary and each claim's text annotated inline as «flagged phrase|CODE», so you can see exactly WHICH occurrence is flagged: - AB!: almost certainly incorrectly begins with 'the' - AB?: may incorrectly begin with 'the' (a close or parenthetical match was found) - AB+: may need 'the' instead of 'a/an' (already introduced in a parent claim) Input: CLAIMS ONLY (no specification or detailed description needed). IMP

Check Literal Support

Checks whether claim phrases and words have literal support in the specification's detailed description. Returns JSON with a summary, a list of unsupported/minimally-supported terms with recitation counts, and each claim annotated inline as «term|LS:N» where N is the number of recitations in the description (0 = no literal support, 1-2 = minimal support). Input: the claims AND the detailed description text. The description is used to verify literal support for the claim language. IMPORTANT: The

Check Profanity

Scans ANY plain text for “patent profanities” (terms a practitioner may wish to avoid, e.g., “critical”, “essential”, “the invention”) and returns JSON listing each term found with its recitation count and surrounding contexts (the flagged term shown as «term»). No specific document structure or headers are required — any text works.

Get Parts List

Extracts the parts list (reference numbers and their names) from the specification's detailed description and returns JSON flagging inconsistent naming, missing parts, and parts found only in the figures, with per-name specification and figure recitation counts. Input: the detailed description text; optionally, plain text extracted from the figures to cross-check specification vs. drawing part naming.

Check Figure Consistency

Cross-checks figure numbers across the brief description of drawings, the detailed description, and (optionally) the figures themselves, returning JSON showing, for each figure, whether it is present in each location (true/false, or 'n/a' if no figure text was supplied), plus an issue summary. Input: the specification text (should include the brief description of drawings and the detailed description); optionally, plain text from the figures.

Get Claim Tree

Resolves any multiply-dependent claims to a single best-fit parent (chosen to minimize antecedent basis issues), returning JSON mapping each dependent claim number to its selected parent claim number. Input: CLAIMS ONLY. IMPORTANT: The claims text MUST begin with a line reading exactly 'CLAIMS:' on its own line, immediately followed by the numbered claims, e.g.: CLAIMS: 1. A system comprising a processor... 2. The system of claim 1, wherein... Without the 'CLAIMS:' header line the text will NOT

Predict 101 Rejection

Predicts the likelihood of a 35 U.S.C. §101 (patent eligibility) rejection for each independent claim, returning JSON per claim: the rejection probability (percent), the five most similar real applications (with their full claim text, similarity, and rejection outcome), and the closest USPTO Patent Eligibility Guidance (PEG) example. Only independent claims are analyzed; dependent claims are filtered automatically. Input: CLAIMS ONLY. Requires Analytics or FullUser group access. IMPORTANT: The

Predict Art Unit

Predicts the most likely USPTO art units and technology groups for the application, returning JSON lists of [art_unit, probability] and [group, probability] pairs. Uses the independent claims and the abstract. Input: the claims AND the abstract. Requires Analytics or FullUser group access. IMPORTANT: The claims text MUST begin with a line reading exactly 'CLAIMS:' on its own line, immediately followed by the numbered claims, e.g.: CLAIMS: 1. A system comprising a processor... 2. The system of c

Generate Spec Support Text

Generates specification support text from the claims, returning JSON with three sections: 'paragraph_claims' (the claims rewritten in paragraph form as specification support language), 'claims_with_part_numbers', and 'abstract_with_part_numbers' (part reference numbers inserted in parentheses). Input: the full specification text (claims, abstract, and detailed description; the detailed description supplies the part reference numbers). IMPORTANT: The claims text MUST begin with a line reading ex

Run Full Report

Runs the complete N-Spec patent proofreading suite on a document and returns a structured JSON report containing a legend for inline issue markers, an overall quality score (0-100), and per-feature findings from all requested analysis modules: antecedent basis, literal support, patent profanities, figure consistency, parts list, claim singularizer, text generation, and — for Analytics/FullUser users — art unit prediction and §101 rejection prediction. Art unit and §101 prediction are silently sk

Search Prosecution Entities

Search for USPTO examiners, art units, groups, or tech centers by name or number. Returns a list of matching entities with their keys, which can then be passed to get_prosecution_stats for detailed data. Examples: - Search by examiner name: 'Ayaan Alam' - Search by art unit number: '2812' - Search by tech center: '2800' - Browse all examiners in an art unit: '1611' with entity_type 'examiner'

Get Prosecution Stats

Retrieve detailed prosecution statistics for a specific USPTO examiner, art unit, group, or tech center. Use search_prosecution_entities first to find the exact entity key if you're unsure of the name/number. Returns data including allowance rates, office action statistics, interview strategy insights, RCE patterns, appeal outcomes, and strategic recommendations. Use the 'fields' parameter to request only specific data sections when you don't need the full dataset (recommended for comparing mu