Sanction Scanner Is Now Available in Claude’s Connector Directory

Sanction Scanner is now a connector in Claude

Compliance analysts have been pasting case details into chat tools for a while now. A name comes back with forty hits, the analyst copies the list into an assistant, asks which of them look plausible, then goes back to the platform and applies the decision by hand. It gets the job done. The assistant is reading a copy rather than the case, though, and none of what gets said in the chat ends up in the file.

Sanction Scanner is now in Anthropic's connector directory for Claude. Claude reads from and writes to the analyst's own account, so the review and the disposition happen in one place, and every change lands in the case record where an auditor can find it.

What the connector does

Once added, the connector sits in the Claude sidebar as Ask SanctionScanner. The analyst asks for things in ordinary language. A request for the most significant name screening cases opened this week comes back as a table with the scan ID, match count, current status and the date each case was opened. From there you can pull the full detail on one scan, set a risk level, change a match status, or attach a memo to the case.

The exchange below is condensed from a live account, with the placeholder names removed.

Analyst: Could you send me the five most significant Name Screening cases that were opened this week?

Claude: Here are the five most significant Name Screening cases opened this week, ranked by match count (which typically indicates higher complexity or risk exposure):

A few things worth noting: scan 130859 stands out with 1,421 matches, far above the rest, which is likely a very common name generating a lot of noise but still warrants a look. Scan 130883 is the only one already assigned and flagged at Risk Level 3, so it is actively being reviewed. Scan 130828 is interesting despite the low match count. Only 4 hits, which on a less common name could indicate a more precise, higher-quality match worth prioritising. Want me to pull the full details on any of these?

Analyst: Could you set the risk level for the case associated with scan ID 130828 to "High," change the match status to "True Positive Approved," and add a memo note stating "Reviewed" for the case?

Claude: I'd like to confirm before making these three changes to scan 130828. Set risk level to High. Set match status to True Positive Approved. Add memo "Reviewed". Shall I go ahead?

Analyst: Yes please.

Claude: All three updates applied successfully to scan 130828.

The notes under the table are the part that saves an analyst time. A scan on 1,421 hits and a scan on 4 need different handling, and one of the five is already assigned to someone else at Risk Level 3. Working that out by eye means opening each case in turn; here it arrives with the list.

Nothing gets written without approval

Claude does not modify a case by itself. It lists the specific changes first, one by one, then waits. Nothing moves until the analyst approves it.

That is on purpose. Under a risk-based approach the disposition of an alert is a judgement made by a named person, and an examiner going through the file will want to know who made it. An assistant that closed alerts quietly would leave behind decisions nobody can account for. Drafting the change and handing it back keeps the judgement with the analyst and takes the clicking out of it.

The connector cannot see anything the analyst cannot see either. It authenticates as you, with whatever permissions your account already carries.

How this differs from connecting an AML database to an AI agent

Several AML providers now expose their data over MCP. Those integrations give an agent read access to sanctions, watchlist and PEP records, which is what you want if you are building your own screening logic and need the underlying data in a machine-readable form.

This one works at the level of the case rather than the record. It exposes case operations, which is a narrower thing than a data feed and aimed at a different job: letting an analyst finish a case without switching windows.

Both have a place. Teams building their own engine need the data feed. The connector is for teams whose cases already sit in Sanction Scanner and who would rather stop copying them into a chat window.

What it runs on

The connector runs on the Sanction Scanner API, using the same data and the same matching logic as the platform. A result in the chat is the result in the dashboard, because there is no separate index or cached copy in between.

Screening through the connector runs on the same coverage as the platform: sanctions, PEP and watchlist data from more than 220 countries, with international lists updated every 15 minutes.

Getting set up

The listing is at claude.ai/directory/connectors/sanction-scanner. Any Claude account can add it, and it authenticates against an existing Sanction Scanner subscription. Setup takes a few minutes, and nobody from your engineering team needs to be involved.

It goes in as a community connector. Anthropic reviews directory submissions for safety, security and compatibility, though a community listing is separate from the verified partner programme.