Book a callLogin

Screen every merchant for real-time risk

The real-time KYB layer that clears the easy merchants before a vendor query or a manual review. Every check returns a sourced verdict, not a black-box score. We co-build it with your team and hand it over to own.

See how it works

Numbers from real applicant pipelines

56%96%

Website & entity coverage

Confidence-gated. It abstains instead of guessing.

~95%

Precision on website identification

Up from about 60% on the prior pipeline.

30–40s

Per merchant, automated

A first pass before an analyst opens the case.

Cover the SMEs others miss

A social page, a marketplace storefront, an app listing. Extruct takes whatever presence a merchant has as the input signal, and it separates look-alikes even when the company name is low-uniqueness. For thin-file SMEs and no-registry markets that open-web read is the primary check, not a supplement.

Any input signal

  • Websiteoptional
  • Social profile
  • Marketplace storefront
  • App store listing
  • Customer reviews

Structured, sourced profile

  • IndustryFreight & logistics
  • Operating inDE, PL
  • High-risk industryNo
  • Reviews4.2 on Trustpilot

Find the real website first

Confidence-gated website and entity discovery, plus a roughly 20-column enrichment stack your analysts and risk models read directly.

  • Right company. Applicant disambiguated from look-alikes at ingestion
  • Abstains, never guesses. Weak evidence returns unverified, not a wrong answer
  • Clean structured fields. Roughly 20 columns your risk models read directly

One engine, every verification scenario

The same engine runs at onboarding, on a schedule for acquiring monitoring, and across freelancer verification, counterparty KYB, and periodic re-screening. It is benchmarked to the Mastercard MMCP program, with each signal a column your team controls.

  • Prohibited goods and MCC. Flagged against jurisdiction rules, with a page-level source
  • Ongoing monitoring. Scheduled re-runs surface only what changed
  • Laundering signal. Declared business versus observed traffic and keywords

Answers you can defend to an auditor

A black-box risk score is a number an officer cannot explain. Extruct returns a verdict on every check instead, and your named policy rules turn those verdicts into a recommendation, with every finding traceable to the source it came from.

  • Calibrated verdicts. Pass, warning, fail, or unverified, by check type
  • Your rules, your call. Verdicts map to clear, review, or escalate
  • Traceable to source. Every finding carries the URLs it came from

A KYB system your team owns

We sit with your team, build the verification around how you actually work a case, then hand it over. The config and the code stay yours.

  • You design the logic. Fields, thresholds, what clears and what routes to review
  • Iterate as risk changes. Adjust it yourself, with no wait on a vendor
  • No lock-in. Own the pipeline; new sources wire in within days

Call the API, or ask in Claude

Same engine, two ways to consume it. Wire it into your pipeline, or run it conversationally. No dashboard to adopt.

REST API

Call it from your onboarding stack

Your app queries Extruct in the backend before a case reaches an analyst. Clean structured fields land ready for your risk models, no new screen for the team to learn.

MCP in Claude

Or ask in Claude, no new dashboard

A compliance officer runs the same checks conversationally inside Claude over MCP, right where the analyst already works.

Prove it on your own applicant pipeline

Read the docs

Co-built and yours to keep$1–2 per company at scaleData residency available