MCP App Store
by KeenEthics
Global Database icon

Global Database MCP Server for Claude and ChatGPT

By Global Database

Overview

Search, analyze, and retrieve corporate data instantly Find companies, access financial data, explore shareholders, and understand full group structures directly through a structured interface. Integrate reliable corporate intelligence into your workflows and turn complex ownership and financial information into clear, actionable insights.

How to connect

  1. Claude: Open the Global Database listing in the Claude connectors directory and choose Connect. Claude listing Global Database exposes 5 tools here.
  2. ChatGPT: Open the Global Database listing in the ChatGPT app directory and choose Connect. ChatGPT listing Global Database exposes 25 tools here.

Tools

check_api_key

ChatGPT
Verify the API connection and return current user info. Returns: id, first_name, last_name, email, is_active, date_joined, last_seen. Call ONLY when the user explicitly asks whether they are connected / who they are, or to diagnose an [AUTH_REQUIRED] error. Never call it speculatively before a normal lookup - it costs a request and adds nothing to the answer.

enrich_employee_contacts

ChatGPT
Find and enrich an employee/contact. Returns full contact details. Matching rules (at least one combination required): - email alone - linkedin alone (person's profile URL, not company page) - first_name + last_name + company_name - first_name + last_name + domain Returns: id, full_name, first_name, last_name, gender, seniority_level, department, function_original_name (job title), phone, email, company_id, company_name, social media. Extended profile fields are requested by default and returned when the account has the Contact Details permission: country, location, headline, work_experience, educations, skills, company_website, company_linkedin. Without the permission they are simply absent. Both email and phone are always returned in a single call. Present only what the user asked for; if they later ask for the other, it is already in this result - do not re-call.

find_people_by_domain

ChatGPT
Find PEOPLE at companies given by WEB DOMAIN, via the contact-enrichment providers. When the user names a domain and asks about people, this is the FIRST call of the turn - do NOT resolve the company first. With a company_id or a company name, use search_employees. At least one filter is required; domain_list is the main one ('vicat.fr'). Narrow with employee_seniorities (static list - see the parameter) and company_locations / employee_locations (plain place names, never ids). Returns: total_employees, total_pages, data: [{id, name, first_name, last_name, job_title, seniority_level, linkedin, country, state, city}]. job_title is the real signal; seniority_level is coarse. The nested company{} block mirrors the PERSON, not the employer - never quote it as company data. A page holds 50 - check total_pages before saying a role is absent. The row 'id' is the provider's, not ours: never pass it to get_employee_details or show it as a link. For email or phone call enrich_employee_contacts with their linkedin.

get_company_by_identifiers

ChatGPT
Look up a company by name, registration number, VAT, ticker, website, email, or LinkedIn. country_code is a HARD filter: a wrong one returns nothing for a company that does exist. By name, pass it only when the user stated the country or it is certain from context (domain TLD, VAT prefix, an earlier turn). NEVER infer it from a legal-form suffix - SRL (RO, MD, IT), SA, GmbH, BV and Ltd span many countries. If the country is unknown, ask the user - do not guess. registration_number, ticker, website, email and linkedin work without country_code. vat_number carries its country in its prefix (FR92057505539 is France) - pass it alone; add country_code only to a VAT with no prefix. Returns the full profile as source-bearing field-groups (basic / address / contact / activity / additional / size), each with its own 'source' - cite the distinct sources. The top-level 'id' feeds get_company_financials, get_company_ownership and get_digital_insights; get_company_details afterwards is redundant.

get_company_by_linkedin

ChatGPT
Look up a company by LinkedIn URL or public ID. Returns a full company profile. Accepts full URLs (linkedin.com/company/global-database) or just the public ID (global-database). Returns the same source-bearing field-group profile as get_company_by_website (each group carries its own 'source'; 'id' is top-level). Calling get_company_details afterwards is redundant.

get_company_by_website

ChatGPT
Look up a company by website URL or domain. Returns a full company profile. Accepts bare domains (tesco.com) or full URLs (https://tesco.com). Returns the full profile as source-bearing field-groups (basic / address / contact / activity / additional / size), each with its own 'source' - cite the distinct sources. The 'id' is top-level. Calling get_company_details afterwards is redundant - this already returns the full profile.

get_company_details

ChatGPT
Get a full company profile by company_id. Returns the same data as get_company_by_website, get_company_by_linkedin, and get_company_by_identifiers. Only useful after prospecting (which returns only id + name) when the user asks for details on a specific result.

get_company_financials

ChatGPT
Get company financial data by company_id. Returns balance sheet, ratios, and key metrics per year. Requires a company_id from a previous lookup. If year is provided, returns data for that year only. Otherwise returns all available years. Returns {data: {<year>: {...}}, source: {...}} - per year: currency, employee numbers, share capital, equity, total assets/liabilities, cash, debt ratios, working capital, current ratio, gearing, and YoY changes. The 'source' names the reporting provider - cite it (it is a real provenance, unlike other commercial tools).

get_company_ownership

ChatGPT
Get company shareholders and corporate group structure by company_id. Returns: shareholders (list with share types, quantities, values) and group_structure (tree with parent/subsidiary hierarchy).

get_digital_insights

ChatGPT
Get a company's digital presence: web traffic, rankings, traffic sources, WHOIS, technologies. Requires a company_id from a previous lookup. Returns: domain, WHOIS, global/country/category ranks, engagement metrics, traffic sources (organic, direct, referral, social, email, paid), geographic breakdown, monthly visit history, technology stack, SEO keywords.

get_employee_details

ChatGPT
Get one person's profile by employee_id (the 'id' of a search_employees row). Returns: id, full_name, first_name, last_name, patronime_name, gender, nationality, seniority_level, department, function_original_name (job title), phone, email, company_id, company_name, linkedin, twitter, facebook. Work experience, education, skills and location are NOT in this response - do not offer them. Call it for the row the user picked. When they ask for several or all rows ("details for all 16", "contacts for everyone in sales") that IS the request: issue one call per row in a single parallel batch and present them together. Only never fan out on your own initiative.

get_nomenclature

ChatGPT
Resolve nomenclature values into the ids that prospecting and kyb_search filters expect. Each item: id, name, parents, has_children, active. Which field goes into the filter - the wrong one silently matches nothing: - location_countries, location_regions: the 'parents' string (e.g. '1869197_1869198') - every other filter: 'id' Read the pre-loaded resources FIRST (gdb://nomenclature/country, /company-status, /kyb-countries); call this tool only for what they miss. On a hierarchical type always pass 'search' instead of paging the tree. nace, isic, industry-focus, sic/*, country and region answer from a local index in a flatter shape: {"source": "index", "matches": [{code, label, exact, ids: {<type>: [value, ...]}}]}. Copy the values under 'ids' straight into the filter - already the right field per nomenclature. 'exact': false means that code does not exist and you got its parent group - say so. Never show nomenclature data to the user; use the ids silently.

kyb_company_details

ChatGPT
Get official registry details for a company. Accepts any company_id (from kyb_search or any company lookup). Returns: id, name, registration_number, vat_number, incorporation_date, status, country_code, address, legal_form, website, linkedin.

kyb_financial

ChatGPT
Get detailed financial statements from official registry across multiple years. Accepts any company_id (from kyb_search or any company lookup). Returns: years (list of reporting dates) and groups - each group has an id, name, and list of financial line items with year-by-year values and progress indicators. Groups: Summary, Income Statement, Balance Sheet, Cashflow, Capital & Reserves, KPIs & Ratios.

kyb_group_structure

ChatGPT
Get corporate group structure from official registry. Accepts any company_id. This is the tool for "subsidiaries of X", "X's group", "companies owned by X" and UBO mapping. That question is TWO calls: resolve the company id first (get_company_by_identifiers with the name and the country the user gave), then call this. Never answer it from your own knowledge - subsidiary names and registration numbers you did not receive from this tool are fabricated. Two modes: - full (default): complete corporate family tree across all levels - lite: immediate parent-subsidiary connections only Defaults to 'full' for the complete hierarchy / UBO mapping. Use 'lite' only when the user just wants the immediate parents and subsidiaries, or to keep the call cheaper. Returns: tree of [{id, name, country, registration_number, selected, children}].

kyb_officers

ChatGPT
Get company directors, secretaries, and officers from official registry. Accepts any company_id (from kyb_search or any company lookup). Paginated. Returns: [{id, first_name, last_name, job_title, appointed_at, date_of_birth_year, address}].

kyb_shareholders

ChatGPT
Get company shareholders from official registry. Accepts any company_id (from kyb_search or any company lookup). Returns: [{id, company_id, name, quantity, currency, share_price, share_type, share_value, total_value}].

kyb_ubo

ChatGPT
Get the Ultimate Beneficial Owners (UBO) of a company: the natural persons who ultimately own or control it, with effective ownership traced through every corporate layer. Accepts any company_id. This is the tool for "who ultimately owns X", "beneficial owners", "UBO". kyb_shareholders shows only the first layer; kyb_group_structure shows companies, not persons. Returns: company, summary {threshold, total_ubos, ubo_determination_conclusive, percentages_are_minimum, ownership_breakdown, enrichment}, data [{full_name, ownership {direct/indirect/effective_percentage}, paths}], terminal_entities (unresolved branches, with reason). enrichment: 'internal' (default) uses Global Database data only. 'external_if_needed' BUYS external ownership reports for unresolved branches - real cost, one enrichment unit per report bought. Use it only after the user agreed to it and to max_enrichment_units (required). If the result is not ready it returns a job_id: follow it with kyb_ubo_job_status.

kyb_ubo_job_status

ChatGPT
Follow a UBO job returned by kyb_ubo (a job_id). Free - costs no credits. Waits briefly for the job; once complete, returns the full UBO result (same shape as kyb_ubo). While still running it returns the status and progress - tell the user and offer to check again; do not call it in a loop within one answer.

prospecting

ChatGPT
Search and filter companies by country, industry, size, revenue and many other criteria. BEFORE any query beyond country + industry + size, call read_reference("prospecting-filters"): exact field names, formats, and which nomenclature field each filter takes. A guessed name or the wrong field returns zero rows, not an error. Never send a blank value - "" or [] is a real filter upstream and matches nothing. Omit the parameter instead. Resolve every id-based filter first, never guess: location_countries/location_regions take the 'parents' string, the classification filters take 'id' (full mapping in the document above). Prefer the pre-loaded gdb://nomenclature/country and /company-status, else get_nomenclature(type, search=...). For "top", "largest", "best" or any ranked query always add sorting_field and sorting_direction. Returns: total_companies, total_pages (max 1000), data: [{id, name}] - minimal rows. COST: each call consumes credits - for a count only, set per_page=1 and read total_companies.

read_reference

ChatGPT
Read one reference document on how to USE the other tools. Returns markdown. Call it before building a non-trivial call, never to decide WHICH tool to use, and never before a lookup you can already make. Costs no credits, returns no company data, and one call per document per conversation is enough. - tool-mechanics: parameter combinations per tool, which nomenclature field goes in which filter, result-shape quirks, pagination limits. - prospecting-filters: the full prospecting filter catalog: exact field names and value formats. Read before any prospecting query beyond country + industry + size. - answer-format: how to lay out each payload shape (profile, financials, officers, group tree, prospecting list, digital insights) and the suggested-action category catalog. - intent-glossary: keyword -> category/tool routing, for a phrasing you cannot place.

search_employees

ChatGPT
Search PEOPLE: the employees of a company, or people by role, department or name. At least one filter is required; company_id answers "who works at X". Search WIDE first: coverage is thin per company (single digits is normal), so send company_id ALONE and read the job titles off the rows. Adding department/seniority/job_title on top of a small company returns 0 for people who ARE listed. department/seniority take nomenclature ids - copy them from the pre-loaded gdb://nomenclature/department and /seniority, never from memory. An empty result carries a 'hint' field - follow it. per_page defaults to 50, page max 10; the UI pages them 10 at a time - never re-call for page 2 to show more. Returns: total_employees, total_pages, data: [{id, name, company_id, company_name}] - those four fields only, no contact details. Expand a row with get_employee_details: the person the user picks, or every row when they ask for all (one parallel batch). For one named person's email or phone use enrich_employee_contacts.

autocomplete

Claude

get_company_by_linkedin

Claude

get_company_by_url

Claude

get_company_financials

Claude

get_company_ownership

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/global-database"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/global-database)