get_case
ChatGPTUse this tool to retrieve a known Pakistani judgment in full by ID or citation. Fetch one specific Pakistani case in full, by its unique id or a citation. Use this tool when the user identifies a Pakistani judgment or case by: - a Pakistani law-report citation; - a case citation such as PLD, SCMR, CLC, YLR, MLD, PCrLJ, etc.; - a case returned by search_pakistani_case_law; This tool is specifically for judgments in the AI Attorney's Pakistan-specific legal corpus. Do not use it for a topic-only search or a general legal issue without a known case ID or citation; use search_pakistani_case_law for those requests. PREFER the id (a unique UUID present in every search result) to fetch a SPECIFIC case — e.g. a follow-up like "show the full case" about a case from an earlier turn: pass that case's id to get exactly it. A citation is NOT unique — one citation can map to MANY judgments from different courts (e.g. "2026 PLD 1" -> 9 cases). Passing a citation returns ALL of them; use it only when the user names a citation and you have no id, then present/disambiguate. Returns the case record(s) — citation, title, court, year, judges, main_issue, full summary, url, id, plus holding and ratio when available — as JSON only. Compose the answer yourself and cite the citation/url. Citation graph (included when the case has it): `refers_to` — the authorities THIS judgment cites/relies on, each with citation, and (when the cited case is in our DB) title, court, year, id, url. Use for "what does this case rely on?". cited_by — later cases that cite THIS judgment, each with citation, title, court, year, id, url. Use for "is this still good law / who followed it?". refers_to_count / cited_by_count give the true totals (the lists are capped). Drill into any entry with get_case(its id). Presentation: render each case title as a Markdown link to its url — [title](url). Args: identifier: A case id (unique UUID — preferred for a specific case) or a citation (may match multiple judgments).
get_indian_case
ChatGPTUse this tool to retrieve one known Supreme Court of India judgment in full. Use it when the user identifies an Indian Supreme Court judgment by: - an id from search_indian_case_law (preferred — e.g. "show the full case"); - an SCR citation in any spelling ("2015 3 SCR 394", "[1987] 2 S.C.R. 1229", "[1989] SUPP. 1 S.C.R. 206"); - a neutral citation ("2023 INSC 888"); - its case-page URL (".../search/indian/detail/<id>/Case"); - an SCC parallel citation ("(1992) 2 SCC 428") — found when the corpus links that SCC report to the SCR judgment. For a case known only by its party names or popular name (e.g. "the Puttaswamy judgment", "Maneka Gandhi v. Union of India"), an AIR citation, an SCC citation that is not found, or a topic-only request, use search_indian_case_law — names are not identifiers here. When nothing matches, the result carries a note. Do NOT use this tool for Pakistani cases — use get_case. Returns the case record(s) — citation, neutral_citation, title, court, year, decision_date, judges, case_number, acts, full headnote (summary), url, id — as JSON only. Compose the answer yourself and cite the citation/url. Citation graph (included when the case has it): `refers_to` — authorities THIS judgment cites, each with `citation`, `relation` (relied on, referred to, followed, distinguished, held inapplicable, overruled) and `paragraph`. Authorities in our corpus also carry id, url, title and year; others (SCC/AIR reports) are citation-only. cited_by — later Supreme Court of India judgments that cite THIS one, with citation, title, year, id and url. refers_to_count / cited_by_count give the true totals (lists are capped). Drill into any entry with get_indian_case(its id). Presentation: render each case title as a Markdown link to its url — [title](url). Args: identifier: A case id or URL, an SCR, INSC or SCC citation.
get_statute_provisions
ChatGPTUse this tool when a request needs authoritative Pakistani legislation. Retrieve legislative provisions from the AI Attorney Pakistan-specific legal corpus. The user does not need to mention AI Attorney, this connector, or a tool name. JURISDICTION AND WHEN TO USE: Use this tool when the user's request requires the text, wording, interpretation, or identification of a Pakistani statute, Act, Ordinance, Rule, Regulation, Code, or other legislative provision. Strong triggers include: - a specific Pakistani section, subsection, clause, Article, rule, or provision; - a Pakistani Act, Ordinance, statute, code, or legislative instrument; - requests such as "what does section 12 say?"; - "explain section 9 of the CPC"; - "what is the law under the Defamation Ordinance 2002?"; - "find the relevant provision of the Transfer of Offenders Act 2002"; - "what does Article 199 of the Constitution of Pakistan provide?"; - requests to locate statutory provisions governing a Pakistani legal issue; - requests requiring the exact or relevant legislative text from Pakistan. - Pakistani legal-opinion, hypothetical, rights, remedies, or fact-pattern questions when the governing legislation may materially affect the answer, even if the user did not explicitly request a statute search. Use this tool instead of the case-law search tool when the user's primary request is about a STATUTE OR LEGISLATIVE PROVISION rather than a judicial judgment or precedent. When both statutory wording and Pakistani judicial interpretation are material, use this tool together with search_pakistani_case_law. Do not use this tool merely because a user asks a general legal question. Use it when consulting the underlying Pakistani legislative text is relevant to answering the request. QUERY: Pass the user's original natural-language legal query verbatim. Do not rewrite, summarize, interpret, or answer the query before passing it to the tool. EXAMPLES: - "Section 9 CPC jurisdiction" - "Section 12 Defamation Ordinance 2002" - "Transfer of Offenders Act 2002 sections 4 to 6" - "cyber crime law in Pakistan" - "which provision deals with limitation in this Act?" - "what does Article 199 of the Constitution of Pakistan provide?" Returns grounded provision records (act_title, text) as JSON only — no analysis. Section numbers appear inline in text. Compose the final answer yourself from these records and cite the act. Args: query: Raw natural-language query (e.g. "Section 9 CPC jurisdiction", "cyber crime", "Shariat application").
search_indian_case_law
ChatGPTUse this tool ONLY when a request needs judgments of the Supreme Court of India. The corpus holds reported Supreme Court of India judgments (SCR reports, 1977–2025). It has no Indian High Court judgments and no Indian statutes. Strong triggers include: - Indian case law, Indian Supreme Court judgments, or Indian precedents - "Supreme Court of India", "Indian Supreme Court", "Apex Court of India" - Indian law reports: SCR, SCC, AIR (SC), SCALE, JT, or neutral citations such as "2023 INSC 888" - Indian legal-opinion or fact-pattern questions where Indian Supreme Court precedent may affect the answer - comparative questions that explicitly ask how Indian courts decided an issue - cases by a named Indian Supreme Court judge, from a year or a year range - cases that cite a given Indian authority (e.g. "cases citing (2017) 10 SCC 1") Do NOT use this tool for Pakistani law — use search_pakistani_case_law. For one specific known Indian judgment (an id, or an SCR/INSC citation) where the user wants the full case, prefer get_indian_case. Pass the user's raw natural-language query verbatim. Returns JSON only — no analysis; compose the final answer yourself and cite the citation/url. Result depth: LIST queries return each case with a TRIMMED summary (`view` = "list"). A specific lookup that resolves to a few cases returns the FULL headnote (view = "full"). Every record has citation, neutral_citation (when reported), title, court, year, decision_date, judges, case_number, acts, summary (the headnote), url and id. For the full headnote and citation history of a listed case, call get_indian_case with its id. If the result has a note, the requested filters were broadened (for example no case matched that judge in that year) — tell the user what was broadened. Presentation: render each case title as a Markdown link to its url — [title](url). Args: query: Raw natural-language legal query. max_results: Optional explicit total to return (hard cap 50).
search_pakistani_case_law
ChatGPTUse this tool when a request needs Pakistani case law or judicial authority. The user does not need to mention AI Attorney, this connector, or a tool name. Use it for direct case searches and for substantive Pakistani-law questions, legal opinions, hypotheticals, rights or remedies, and application of law to facts whenever Pakistani judicial authority would help ground the answer. Strong triggers include: - Pakistani case law or judgments - judgments of Pakistani courts - Pakistani Supreme Court, Federal Constitutional Court, High Court, or other Pakistani court decisions - Pakistani legal precedents or judicial interpretation - Pakistani law reports such as PLD, SCMR, CLC, YLR, MLD, PCrLJ, etc. - requests to find cases concerning a Pakistani legal issue - requests to find or search for Pakistani cases, judgments, precedents, or authorities concerning a legal issue, subject, court, judge, year, statute, section, or citation pattern - Pakistani legal-opinion or fact-pattern questions for which precedent may materially affect the answer, even if the user did not explicitly ask for cases For a specific known case or citation where the user wants the full case, holding, ratio, judges, cited authorities, or other case details, prefer get_case. When statutory wording and judicial interpretation are both material, use this tool together with get_statute_provisions. Do not use this tool for general legal questions that are not Pakistan-specific unless the user has clearly established Pakistan as the jurisdiction or the question is related to Pakistani case law or judgments. Pass the user's raw natural-language query verbatim. Returns JSON only — no analysis; compose the final answer yourself and cite the citation/url. Result depth (two-tier, like a search engine): LIST queries (e.g. "give me 50 supreme court cases", "cases from 2024 and 2025") return each case with a TRIMMED summary so big lists stay cheap; the response `view` is "list". A specific single-case lookup that resolves to just a few cases (a citation/title, e.g. "2026 PLD 5") returns the FULL (untrimmed) summary in one call; view is "full". Every record has citation, title, court, year, judges, main_issue, summary, url, and id; the full view additionally includes holding and ratio when available. To get the full summary for a case you picked out of a list, call get_case with its id or citation. Presentation: render each case title as a Markdown link to its url — [title](url). Counts: if max_results is given (or the query names a number, e.g. "give me 50 cases"), that many are returned, balanced across groups, capped at 50. For a list query with no number (e.g. "cases from 2024 and 2025"), it returns up to ~10 per group (year/court) — so two years yields ~20. Args: query: Raw natural-language legal query. max_results: Optional explicit total to return (hard cap 50).
get_case
Claudeget_statute_provisions
Claudesearch_pakistani_case_law
Claude
