Fused intelligence you can defend.
Dark-web context, on-chain provenance, and identity — correlated into one court-defensible graph. With published error rates.
To investigate money that moved through the dark web, you need three products.
One traces the chain. One has the darknet corpus. One does infrastructure and identity.
Together they cost more than €300,000 a year. None of them fuses with the others. None fuses with your own data.
So an analyst on your team becomes the integration layer — and that is the most expensive way there is to get an answer.
We are the correlation.
One graph. One query. From a darknet marketplace listing, to the wallet that paid for it, to the off-ramp that cashed it out, to the entity behind the off-ramp.
No tool switching. No re-keying identifiers. The correlation is already done.
Four things nobody else has together
We run the nodes
First-party ground truth from twelve-plus blockchain full nodes we operate ourselves. Block heights in your audit record, not a resold vendor API.
We publish our error rates
Measured against a closed-loop chain where we hold the answer key, plus honeytoken assets that produce confirmed evidence. Ask any other vendor for theirs.
Glass box, always
Every score ships per-feature contributions. Every alert cites the FATF or FinCEN rule. Every brief cites the rows it came from.
Civil liberties, audited
We found inferred ethnicity in our own product and published what we did: 4,059,250 rows deleted, 13,801 already-served scores withdrawn. The producers sit behind a gate that fails closed.
Don't take our word for any of it
We publish six datasets, free and unrestricted, built from public-domain government records and from a Bitcoin archival node we operate ourselves. Every summary figure in them is a live formula pointing back at the rows behind it, so you can click any number and trace it.
They include the places our own work is weakest. One workbook documents a defect in our own CoinJoin detector and publishes the filtered subset we would actually stand behind, rather than the flattering headline number.
Bring us a case you have already worked.
Not a demo on our data — a real matter where you already know the answer. It is a better test than anything we could show you.