Use case · VerifyDesk
Screen a company or person against sanctions lists from an agent
Screen a person or company name against the OFAC SDN and Consolidated, UN, EU and UK sanctions lists in one call, aliases included, with scored matches and the date of each list.
The job
Before paying a supplier, onboarding a customer or shipping to a new buyer, a business screens the name against the sanctions lists that apply to it. The US Treasury publishes the OFAC SDN list and its Consolidated (non-SDN) list, the UN Security Council its Consolidated List, the European Commission the EU consolidated financial sanctions list, and the UK Foreign Office the UK Sanctions List. Names appear with aliases and in several transliterations, which is why an exact string search is not enough.
The useful output is a short list of candidate matches with a similarity score, the list and program each comes from, and proof of which edition of each list was checked.
Why an agent needs a tool for it
An agent that moves money or onboards counterparties cannot stop and ask a human to run a search on three government websites. It needs one call that covers the lists, matches aliases and spelling variants, and returns structured candidates it can escalate.
screen_sanctions always reports how many entries were screened, which lists answered and data_as_of, the oldest refresh among them. Keeping that with the decision is what makes a screening record checkable months later.
The calls
2 tool calls.
{
"name": "screen_sanctions",
"arguments": {
"name": "CONGOMET TRADING HOUSE"
}
}{
"query": "CONGOMET TRADING HOUSE",
"screened_entries": 85027,
"screened_parties": 33554,
"lists": [
"EU Consolidated",
"OFAC Consolidated (non-SDN)",
"OFAC SDN",
"UK Sanctions List",
"UN Consolidated"
],
"lists_unavailable": [],
"data_as_of": "2026-10-11T23:10:28.163Z",
"stale": false,
"matching": {
"min_score": 0.85,
"version": "2026-10-10.4"
},
"match_count": 3,
"matches": [
{
"name": "CONGOMET TRADING HOUSE",
"type": "entity",
"programs": "DRC",
"list": "UN Consolidated",
"score": 1,
"match_type": "exact",
"source_id": "CDe.004"
},
{
"name": "Congomet Trading House",
"type": "entity",
"programs": "COD",
"list": "EU Consolidated",
"score": 1,
"match_type": "exact",
"source_id": "EU.3672.18"
},
{
"name": "CONGOMET TRADING HOUSE",
"type": "entity",
"programs": "The Democratic Republic of the Congo (Sanctions) (EU Exit) Regulations 2019",
"list": "UK Sanctions List",
"score": 1,
"match_type": "exact",
"source_id": "DRC0018"
}
],
"notice": "A match is a signal for enhanced due diligence, not a legal determination."
}A listed entity, found on the UN, EU and UK lists, each with its own reference number. This is the case an agent must stop on and hand to a person. Shortened for this page: list_freshness omitted: one entry per list with refreshedAt (our mirror), publishedAt (the publisher's own date), entries and parties.
{
"name": "screen_sanctions",
"arguments": {
"name": "Danone"
}
}{
"query": "Danone",
"screened_entries": 85027,
"screened_parties": 33554,
"lists": [
"EU Consolidated",
"OFAC Consolidated (non-SDN)",
"OFAC SDN",
"UK Sanctions List",
"UN Consolidated"
],
"lists_unavailable": [],
"data_as_of": "2026-10-11T23:10:28.163Z",
"stale": false,
"match_count": 0,
"matches": []
}A clean result still says what was screened and against which edition of the lists, which is the part an auditor asks for. Shortened for this page: list_freshness, matching and notice omitted; they are the same as in the call above.
Limits worth knowing
- A match is a signal for enhanced due diligence, never a legal determination. Similar names collide, and the score is a similarity measure, not a probability of identity.
- Coverage is the OFAC SDN and Consolidated lists, the UN Consolidated List, the EU consolidated list and the UK Sanctions List. National lists of other countries are not screened, so check which lists your obligations name.
- Matching compares names only, in Latin script: dates of birth, nationalities, addresses and vessel identifiers are not compared. Measured on 2026-10-10, 92.36% of 9,130 realistic spelling variants of listed names were found and 0.78% of 36,070 names that are not designated were wrongly flagged (method in the benchmark linked from the product page).
- When the list data is older than the freshness threshold the answer says stale: true. An agent should treat a stale screen as unfinished rather than clean.
Pricing
2 credits for the calls on this page.
| Tool | Credits per call |
|---|---|
| screen_sanctions | 1 |
Credits are prepaid and shared across every server. Starter is $10 for 1,000 credits, packs never expire, and 100 credits are free without a card. Full pricing.
Install
Connect VerifyDesk to your agent.
A remote MCP server over Streamable HTTP. Get a free key, swap it in, and the calls above work as written, with answers reflecting the data on the day you call.
https://verifydesk.agenttoolworks.com/mcp
claude mcp add --transport http verifydesk \ https://verifydesk.agenttoolworks.com/mcp \ --header "Authorization: Bearer atw_live_your_key"
Setup for Claude Desktop, Cursor and Windsurf is in the docs.
Read next
- VerifyDesk: every tool, the data sources and measured reliability.
- Verifying a company from an AI agent, the long-form guide.
- Verify an EU VAT number via API, from an agent
- Validate an IBAN via API before a payment goes out
- Look up a French company by SIREN or name, via API
- Supplier onboarding checks an agent can run on its own
- All use cases
Public data only, from the sources named on the product page. Not affiliated with any registry, list publisher, job board provider or store platform named here.