Sanctions Screening MCP Server

Sanctions Screening

Screen persons, entities, vessels, and aircraft against 25+ government sanctions and watchlists aggregated from public sources. Updated daily. Includes an interactive app for searching and triaging matches with AI assistance.

Interactive App

An interactive screening dashboard is available. Match results display as cards showing source list, entity type, country, DOB, nationality, aliases, and sanctions programs. Clicking a match card triggers AI analysis — true/false positive assessment, recommended next steps, and related entity checks.

Data Sources

• OFAC SDN & Non-SDN (US Treasury) — Specially Designated Nationals, Sectoral Sanctions
• EU Financial Sanctions — Full EU consolidated list
• UK FCDO — UK sanctions list
• UN Security Council — Consolidated UN sanctions
• Canada SEMA — Canadian sanctions under SEMA
• Switzerland SECO — Swiss sanctions list
• Australia DFAT — Australian consolidated list
• Interpol Red Notices — International arrest warrants
• PEP — Politically Exposed Persons (from Wikidata)
• Mexico SAT — Mexican tax authority sanctions
• Belgium, Netherlands, Sri Lanka, Israel — National lists

Capabilities

Name screening:

• Fuzzy match using Jaro-Winkler, Levenshtein, token matching, and phonetic algorithms
• Composite scoring with DOB, country, and identifier adjustments (pass optional identifiers on the screen call)
• Optional tuning: token_aggregate (mean | min | harmonic | max), skip_generic_tokens, include_token_sort, include_entity_levenshtein, or a screening_preset (kyb_strict | sanctions_recall) that overrides those knobs for the call
• Cross-source deduplication (same person listed in multiple sources)
• Entity type filtering (individual, entity, vessel, aircraft)
• Source list filtering

Identifier search:

• Search by passport number, national ID, tax ID, registration number
• IMO number, call sign, MMSI for vessels
• Crypto wallet addresses

Batch screening:

• Screen up to 50 names in a single call
• Batch limit_per_case (default 10, max 50) caps matches per case; per-case identifiers and scoring overrides (screening_preset, token_aggregate, etc.) are supported

Workflows

Quick triage: Screen a name, review match cards showing source list, country, DOB, programs, and alias count. Analyze whether each hit is a true or false positive and recommend next steps.

KYC onboarding: Screen a customer name with DOB and country. Review any matches above the threshold. Use identifier search to confirm or rule out matches.

Vessel due diligence: Search by IMO number to find sanctioned vessels. Cross-reference with entity screening for the vessel operator.

Bulk compliance: Batch-screen a list of counterparties, vendors, or beneficiaries. Set batch defaults for screening_preset, token_aggregate, skip_generic_tokens, include_entity_levenshtein, or override per case. Flag any with matches for manual review.

Scoring

Scores range from 0 to 1. Default threshold is 0.85; raise it (e.g. 0.90–0.95) for entity KYB when the name is generic. Lists include PEP and registries — not every hit is a sanctions designation.

• 0.95+: Very strong match — likely the same person/entity
• 0.90-0.95: Strong match — high confidence, review recommended
• 0.85-0.90: Moderate match — warrants investigation
• Below 0.85: Weak match — likely a different person

Token options (name screen): token_aggregate=mean (default) averages per-token similarity; min requires every counted token to align well (fewer false positives on “Trading LLC”-style names); harmonic is between mean and min; max is the most lenient. skip_generic_tokens=true drops legal/shell tokens from that path. include_token_sort=false disables token-order fuzzy matching.

Secondary adjustments: DOB match (+0.10), year match (+0.05), DOB mismatch (-0.15), country match (+0.05), identifier match (+0.15).