convert_file_to_markdown
ChatGPTConvert a file from a URL to Markdown. Handles almost any document type (PDF, Word/DOCX, PowerPoint/PPTX, Excel/XLSX, HTML, CSV, JSON, XML, EPub, images, …) via the markitdown engine. For scanned PDFs that have no extractable text, it falls back to Google Vision OCR. For security the URL must point to a trusted legal source (e.g. justice.cz, gov.cz, e-sbirka.cz, the courts, europa.eu, echr); other hosts are rejected. Company document workflow: search_company → search_company_documents → convert_file_to_markdown(detail_url) Pass the detail_url from search_company_documents directly — the backend handles the stateful session required to download PDFs from or.justice.cz.
currency_conversion
ChatGPTConvert a monetary amount between two currencies using the official central-bank reference (fixing) rate for the user's jurisdiction. The provider is chosen automatically from the jurisdiction: CNB (Czech National Bank) for Czechia, NBP (National Bank of Poland) for Poland, and the ECB (European Central Bank) for the Eurozone and every other currency. Cross-rates (e.g. USD→HUF) are computed via the provider's base currency. This is the central-bank reference rate, not a bank/transaction rate.
get_current_time
ChatGPTGet the current date and time, including the day of the week. Use this whenever you need to know today's date or the current time — for example to compute filing deadlines or limitation periods, to judge whether something is currently in force, or to answer "what is the date/time now". The time is returned in the user's jurisdiction timezone and the weekday/date is also given localized to the jurisdiction's language.
get_document
ChatGPTGet complete document content by UUID(s). Accepts one or more UUIDs.
get_entitlements
ChatGPTReport the current account's entitlements and the tools available to it. Use this to find out who you are authenticated as, which country/market and subscription tier your account resolves to, and exactly which tools you are allowed to call. Takes no arguments. Call it whenever the available tool list looks empty or unexpectedly small: it distinguishes an account with no active subscription or no resolved market from an authentication problem, which otherwise look identical in the client. Returns a JSON object with: authenticated: whether a verified WorkOS session is present. user_id, email: identity (empty when anonymous). tier: effective subscription tier (basic | pro | premium | enterprise). tier_source: where the tier came from (admin-override | store | jwt | default). country: the active market driving tool/collection availability. country_source: where the market came from (admin-override | subscription | profile | jwt | deployment-default). subscription_status: the store row's subscription status, if any. purchased_jurisdictions: markets granted by an active purchase (Stripe product features) — these FORCE the country while entitled. profile_jurisdiction: the user's own market choice (used when no entitled purchase, or to pick within a multi-market one). entitlements: raw entitlement slugs from the JWT (usually empty on the connector — the OAuth flow carries no org context). available_tools: tool names visible to this account. fallback_country: deployment default (APP_LANGUAGE) used when nothing above yields a market.
get_law_detail
ChatGPTRetrieve the full text of legal fragments (Paragraph § or Article Čl.) from the Czech Collection of Laws (e-Sbírka). Accepts an array of law references and returns all results. If user has content with references like "podle § 2 89/2012 Sb.", pass all referenced laws in a single call.
get_law_provision
ChatGPTGet the exact text of ONE cited provision of a law and check that it exists. Use it to verify every provision you cite before you quote it or rely on it. Pass the citation as people write it, with the act named by its number, abbreviation or title: "§ 1 odst. 1 zákona č. 89/2012 Sb." / "§ 2910 občanského zákoníku" "§ 823 BGB" / "artigo 31.º do Decreto-Lei n.º 555/99" "Art. 6 Abs. 1 lit. f DSGVO" / "Article 6(1)(f) GDPR" / "§ 77 Väärtpaberituru seadus" Returns exists (true/false), act_key, node_key, cite_path, text (the wording in force on date), valid_from, valid_to, heading, title and source_url. When the provision does not exist, exists is false with reason ("node_not_found", "act_not_found", "unparsed") and nearest valid paths of the same act.
legislation_changes
ChatGPTList Czech Parliament bills/prints (sněmovní tisky) that changed within a date window, classified by outcome — did each bill become law or not? Use for questions like "what legislation passed last month", "which bills were rejected in Q1", or "what changed in Parliament between two dates". For "what is the current state of bill X" or a rolling "last N days" window, use `track_legislation` instead. Each result carries the outcome ('enacted' = became law / 'not_passed' = rejected, withdrawn or not accepted / 'in_progress' = still being deliberated), the raw status, the Collection-of-Laws reference ('sbirka_ref', e.g. "35/2026 Sb.") when enacted, the dated events inside the window, a link to the psp.cz timeline, and the bill's source 'documents' (the full print, which embeds the explanatory memorandum / důvodová zpráva).
search_caselaw_parallel
ChatGPTSearch case law (court decisions, judgments, rulings). Use this tool to find how courts interpreted the law in specific cases.
search_company
ChatGPTSearch Czech government registries for a company by IČO (identification number) or by name (nazev). When searching by IČO: Returns combined data from: - Obchodní rejstřík (business registry) — company extract with name, address, executives, shareholders - Insolvenční rejstřík (insolvency registry) — any insolvency proceedings - Živnostenský rejstřík (trade licensing registry) — trade licenses When searching by name (nazev): Returns list of matching companies from the business registry. Use the IČO from results to get full registry data. At least one of ico or nazev must be provided. Use this when the user asks about a Czech company, its registration, insolvency status, or trade licenses.
search_company_documents
ChatGPTList documents from Sbírka listin (document collection) for a Czech company by IČO. Returns list of filed documents (annual reports, founding documents, etc.) with their types, dates, and detail URLs. Workflow: search_company (get IČO) → search_company_documents (list docs) → convert_file_to_markdown (read file). Use the detail_url from results to access document URLs.
search_court_hearings
ChatGPTFind upcoming hearings (nařízená jednání) of Czech district, regional and high courts from the official schedule InfoJednání (infosoud.gov.cz) — today up to about a month ahead. Use for questions like "when is the next hearing in case 15 C 123/2026 at Okresní soud v Písku", "what does judge Novák have on Monday", or "hearings at Městský soud v Praze on 12 November". At least one of court, case_number, judge or query is required. Each hearing carries date, time, courtroom, case number (spisová značka), judge (řešitel), hearing type, cancelled / non-public flags, result, 'checked_at' (when DirectCase last confirmed it at the source) and 'source_url' (the official hearing detail).
search_identifier
ChatGPTSearch for case law documents by identifiers (e.g. file number/spzn/cislo jednaci/spisova znacka/cislo rozhodnuti). Accepts an array of identifiers and returns merged results.
search_keyword
ChatGPTKeyword search across legal document collections using Elasticsearch. Returns matched documents with highlighted text snippets. Use this for precise term matching — exact phrases, specific legal terminology, law numbers, or known keywords. Does NOT use semantic/vector similarity. For semantic (meaning-based) search, use search_similar instead. For combined keyword+semantic with AI summaries, use search_caselaw_parallel / search_law_parallel.
search_law_parallel
ChatGPTSearch legislation (laws, acts, regulations, directives). Use this tool to find statutory law, regulations, and legal codes.
search_law_text
ChatGPTFind provisions of a country's law by their words and get the exact path of each. Use it when you know what a provision says but not where it is; to read or verify one cited provision use get_law_provision. Pass the words of the provision, not a question: "kdo jiného úmyslně usmrtí" / "výpověď z pracovního poměru z důvodu nadbytečnosti" A citation ("§ 2910 občanského zákoníku") also works: the cited provision comes first (cited: true). Returns hits in order of relevance. Each hit has citation ("§ 140 odst. 1 40/2009 Sb."), act (zakon, title), provision (the path as parameters: paragraf, odstavec, pismeno, bod …), node_key, matched (the words found, marked <em>), text (the wording valid on date), valid_from, valid_to, provision_url (the laws-api call that returns the full wording with its parent units) and verified (laws-api confirms the path exists).
search_regulatory_parallel
ChatGPTSearch regulatory body decisions and guidelines. Use this tool to find decisions from administrative and regulatory authorities (e.g. energy, competition, financial and similar regulators).
search_similar
ChatGPTSemantic search for similar legal documents using Qdrant vector search. Finds documents by meaning similarity, not exact keyword matching. Use this when: the user describes a legal situation in natural language, wants to find similar cases/laws, or keyword search doesn't capture the intent. For exact keyword matching, use search_keyword instead. For combined keyword+semantic with AI summaries, use search_caselaw_parallel / search_law_parallel.
track_legislation
ChatGPTTrack the legislative process of Czech Parliament bills/prints (sněmovní tisky). Filter by current state and/or recent activity to find which bills are at a given stage or changed recently. Use for questions like "which bills are in the 2nd reading", "what changed in Parliament in the last 7 days", or "which prints were approved recently". Each result carries the current status, the dated process timeline, and a link to the live psp.cz history page.