Type a company registration number. Five independent sources query in parallel. A deterministic scoring engine produces a verdict. Output: a sourced dossier ready for the MLRO queue.
All five run concurrently. Total wall-clock to verdict is bounded by the slowest source, not the sum.
Companies House (UK), Brønnøysundregistrene (NO), CVR (DK), Äriregister (EE), INPI RNE (FR), GLEIF LEI. Officers, status, incorporation date, registered address, declared beneficial owners.
EU FSF, OFAC SDN, UN Security Council, UK HMT, Swiss SECO consolidated lists — daily-refreshed, fuzzy-matched with corroboration by DOB / country / identifier. OpenSanctions PEP screen (747k entities) against every declared officer + UBO.
Live news search via Brave, LLM-filtered for subject relevance, ranked by Tier 1 / 2 / 3 source quality. Each hit carries a severity tag and a temporal-decay factor — a fine from 2016 weighs less than a fine from last quarter.
Each declared director cross-validated against the corporate leadership page (Playwright-rendered with anti-bot evasion), Brave LinkedIn search with role + nationality disambiguation, and OpenSanctions for PEP overlap. 0–100 score per director with sources cited.
Domain age via RDAP / WHOIS, DNS reachability + TLS validity, MX presence, Wayback first-snapshot age, GLEIF LEI status, LinkedIn company-page detection.
Every signal in every linked case is traceable to its authoritative source.
Fictional 3-week-old UK Ltd. Declared 75% UBO is a real HMT-listed individual. Exact-match override forces RED. Every shell-pattern signal fires.
Real UK luxury PLC, public sources. System correctly approves — clean across sanctions, PEP, and adverse media despite 9 directors to verify.
An ordinary 4-year-old UK distributor — clean entity, real web presence, no adverse media. One director’s name is a romanization variant of a Swiss SECO-listed individual: exact-match screening clears it, the transliteration matcher flags AMBER for enhanced due diligence.
The verdict is rule-based. Same input + same lists = same verdict, every time. Versioned, MLRO-approved scoring weights (v1.2.0). FATF-aligned 5-category typology. RED requires either a hard-override trigger (exact sanctions, dissolved entity) or score ≥ 50 with corroboration from at least 2 non-zero categories. Single-category accumulation tops out at AMBER.
Regulated-entity attenuation. When the counterparty is itself a licensed financial institution (bank, investment firm, payment institution), reputational signals from enforcement findings are attenuated 0.5×. Mirrors how real MLROs already discount routine-supervision noise for incumbents.
LLM commentary is advisory. An LLM-drafted paragraph in senior-MLRO voice sits on top of the verdict — agrees / disagrees, flags context the rules can't see, suggests reviewer focus. Clearly labelled. The deterministic verdict remains the source of truth.
Data is local. Sanctions + PEP datasets are mirrored on disk, refreshed nightly from OpenSanctions. No API roundtrip on screening — fast, free, audit-trail-preserving. The 48-hour staleness gate refuses to trust expired data.
UK, NO, or DK today. EE + FR live as soon as the keys arrive. A sourced dossier in under two minutes.
Start a screening