OpenRegistry

list_jurisdictions

Per-country reference. In-process metadata lookup — no upstream HTTP, no rate cost. Pass EXACTLY ONE of: • jurisdiction='GB' — registry name + URL, data license, company-ID format with examples, native-to-unified status enum mapping, and the list of tools supported. • supports_tool='get_officers' — which jurisdictions implement a given tool (name matched exactly against registered tool names). Use this to plan calls. Do NOT use it to fetch actual data — call the named tool directly. Passing both params or neither returns a 400 with the matching usage hint.

search_companies

Search a national company registry by name or keyword. Pass EXACTLY ONE of: • jurisdiction='GB' - single country, direct. • jurisdictions=['GB','NO','FR'] - multi-country when you're unsure; the server asks the user to confirm (clients with MCP elicitation) or errors back asking you to ask in chat. Per-tier cap on distinct countries per call: anonymous=3, pro=10, max=30, enterprise=unlimited. Returns candidates with unified fields (company_id, company_name, status, incorporation_date, registered_address) plus raw upstream jurisdiction_data. For country-specific filters (FR ca_min, CZ czNace, CH canton, etc.) pass the filters object — call list_jurisdictions for the per-country schema.

search_officers

Find people holding or who held officer positions (director, secretary, member, partner) in a jurisdiction's registry by name. Returns candidates with officer_id, name, and (where exposed) appointment count. Entry point for person-centric investigations. Use this when only the person's name is known. If you already have the company_id, call get_officers instead — it returns the authoritative board roster for that company; this tool is a name-search index and may miss officers whose names are redacted, transliterated, or stored under a former spelling upstream. Read-only — no upstream mutation, results briefly cached. Name-matching semantics vary per registry (exact / prefix / token / fuzzy); full names work best, partial names broaden recall. query matches against the registry's officer name index only — not a free-text web search — so role keywords, synonyms, or company names won't match. limit is an upper bound capped at 100; the registry may return fewer. Unsupported jurisdictions return 501; call list_jurisdictions({supports_tool:'search_officers'}) for the coverage matrix.

get_company_profile

Fetch a company's structured profile by registry-specific ID. Returns unified top-level fields (company_id, company_name, status, status_detail, incorporation_date, registered_address) plus raw upstream under jurisdiction_data. status is a coarse active/inactive/dissolved/unknown enum; status_detail keeps the registry's native string; registered_address is flattened to one string with the upstream nested form preserved in jurisdiction_data. Does NOT bundle officers / shareholders / filings / charges — call those tools separately. ID format varies per registry; pull company_id from search_companies rather than guessing. For per-country ID format and the full jurisdiction_data field catalogue, call list_jurisdictions({jurisdiction:'<CC'}). include.vr (CZ-only), include.history (PL-only), and include.establishments (BE-only) each trigger an EXTRA upstream call and are silently ignored elsewhere — only set the ones you actually need. fresh=true bypasses the short-TTL cache and forces a registry round-trip; default cached path is fine for most lookups since profiles change only on filings.

list_filings

Return a company's filing history, newest first. Each filing has filing_id, filing_date, category, description, and (when upstream exposes one) a document_id that round-trips to get_document_metadata / fetch_document. Raw upstream fields preserved under jurisdiction_data. Filter via the optional category. Common normalized values: 'accounts', 'annual-return', 'capital', 'charges', 'confirmation-statement', 'incorporation', 'insolvency', 'liquidation', 'mortgage', 'officers', 'resolution'. Native upstream form codes also accepted. This tool returns metadata only — call fetch_document on document_id for the actual filing bytes. has_document=false means the body is paywalled or unavailable upstream. Pagination uses limit (default 25, max 1000) plus cursor (GB) or offset (IE). Unsupported jurisdictions return 501; call list_jurisdictions for per-country category values and pagination style.

get_shareholders

Return the shareholders / members / quota-holders — the legal-statutory equity roster published by the registry, no ownership-threshold filter. Use for any shareholder / member / quota-holder question. Shareholders are a DIFFERENT concept from beneficial owners (PSC / UBO), who appear on a separate register only when above a statutory control threshold (typically 25%). The two can disagree: a 10% shareholder is on the members register but not the PSC register; a corporate trustee can be a PSC without appearing on the members register. Disclosure is legal-form-conditional: private-limited / LLC forms typically expose quota-holders publicly; joint-stock / public-limited forms keep shareholders in a private book. Responses therefore may be an empty list, a pointer to the relevant filing (call fetch_document on the returned document_id), or a statutory explanation — every response carries a disclosure flag and/or note. Raw upstream in jurisdiction_data. Unsupported jurisdictions return 501; call list_jurisdictions({supports_tool:'get_shareholders'}) for the coverage matrix. company_id must come from search_companies (do not synthesize); fresh=true bypasses cache and refetches from upstream — useful only right after a known filing event since the roster changes only when filings are submitted.

get_officers

Return a company's officers — current directors, secretaries, members, partners, board members, procurists, liquidators, plus historical resignations by default. Each officer has a unified shape (officer_id, name, role, appointed_on, resigned_on, is_active) plus raw upstream in jurisdiction_data. Limitations: • Role labels pass through in the registry's native language (Styremedlem, Předseda představenstva, Président) — translate client-side. • Birth-date precision varies by registry. • Officer-ID stability varies: corporate officers usually carry the corporate's own company_id; natural persons may carry a synthetic index. • Some registries mask names under GDPR (masking is upstream). • Jurisdictions without an officer feed return 501. When NOT to use: if you only have a person's name (no company_id), call search_officers first to find candidates. For shareholders / quota-holders, use get_shareholders — separate register, different concept. include_resigned=false narrows to currently-serving only (some registries don't expose resigned officers regardless of this flag). group_by_person=true is CZ-only and collapses one person's consecutive appointments (member → chair → vice-chair) into a single entry with the chain under jurisdiction_data._appointments[]. fresh=true bypasses the short-TTL cache.

get_document_metadata

Retrieve metadata for a filing document by document_id (from list_filings). Returns available content formats with byte sizes, page count, source URL, creation date. Raw upstream preserved under jurisdiction_data. Call this before fetch_document when a document may be large or its format is unknown — cheaper than a full fetch. Do NOT construct or guess document_id — some registries use composite IDs that must come from list_filings; synthesized IDs will 404. Empty available_formats means the body is paywalled or unavailable upstream. Unsupported jurisdictions return 501; call list_jurisdictions({supports_tool:'get_document_metadata'}) for the coverage matrix. fresh=true bypasses cache but is rarely useful — filings are immutable once registered.

fetch_document

Read a filing's content by document_id from list_filings. Numbers and prose live inside the document; list_filings metadata only locates filings. RESPONSE SHAPES: • kind='embedded' (under max_bytes, ~20 MB default) — full bytes_base64, source_url_official (evergreen registry URL), source_url_direct (short-TTL signed proxy URL). PDFs render as a native document block. • kind='resource_link' (oversized) — NO bytes_base64. Returns reason, next_steps, both source URLs, and index_preview {page_count, text_layer, outline_present}. Call get_document_navigation to locate pages, then re-call this tool with pages='N-M' and format='pdf'|'text'|'png'. When NOT to use: to enumerate a company's filings, use list_filings. To check size or available formats before deciding, use get_document_metadata. Never synthesize document_id — composite IDs will 404. CRITICAL: on failure (rate limit / 5xx / timeout) do NOT fabricate names, numbers, or dates — tell the user what failed and offer retry or source_url_official. Outline titles, previews, and navigation snippets are for LOCATING pages, never for quoting. max_bytes is a hard inline cutoff: raising it forces full proxy/R2 transfer (slower, costlier); the default returns resource_link for big PDFs so you can page-fetch. fresh=true bypasses the R2 cache and refetches from upstream — filings are immutable so it's rarely needed. source_url_official auto-resolves from the most recent list_filings call; company_id / transaction_id / filing_type / filing_description are overrides only when document_id did NOT come through list_filings.

get_document_navigation

Return a navigation index for a cached document: PDF outline / bookmarks, per-page text previews (~200 chars each), keyword-matched landmarks (balance sheet / directors report / auditor report), text-layer classification, and source URLs. Call this FIRST for PDFs too large to embed (fetch_document returned kind='resource_link'). Use outline / previews / landmarks to pick a page range, then re-call fetch_document with pages='N-M' for the authoritative content. company_id / transaction_id are overrides only for the rare case where document_id did NOT come through list_filings (otherwise auto-resolved from the side-cache). Navigation aids only: previews, outline titles, landmark matches, and snippets may be truncated or contain OCR errors. NEVER cite them for figures, quotes, dates, or names — always quote from a subsequent fetch_document page-range fetch. Requires the document bytes to already be cached via fetch_document; fresh=true re-runs pdfjs against the cached bytes but does NOT re-fetch upstream, so use it only when the index looks stale or empty.