CRYPTO/OR/NOT

Crypto exposure intelligence for fiat rails

Crypto flow is on your rails.Your monitoring cannot see it.

Crypto or Not checks the counterparties in your wires, ACH, and card flows against a registry of crypto businesses. Every match comes with the public record behind it. Built for BSA and AML teams.

Run a live assessmentRead the API docs

Exposure by counterparty

3 of 3

  • ATM OPS INCachCrypto0.98
    Matched observable
    ATM Ops Inc
    Type
    legal_name
    Entity
    Bitstop
    Category
    Kiosk Operator
    Source
    msb.fincen.gov
    Observed
    Aug 8, 2026 · reviewed
  • FORIS DAX INCwireCrypto0.97
    Matched observable
    Foris DAX, Inc.
    Type
    legal_name
    Entity
    Crypto.com (US)
    Category
    Exchange, Broker
    Source
    crypto.com
    Observed
    Aug 8, 2026 · reviewed
  • BLOCK INCwireReview0.60
    Matched observable
    Block, Inc.
    Type
    legal_name
    Entity
    Cash App (Block)
    Category
    Broker
    Source
    dfs.ny.gov
    Observed
    Sep 12, 2026 · reviewed
Real registry entry. Real classifier output. Confidence calibrated on synthetic descriptors.

Built from public records, with the source on every row

crypto businesses in the registry
127

crypto businesses in the registry

names that appear in payment text
345

names that appear in payment text

of those names reviewed against their source
93%

of those names reviewed against their source

distinct public sources cited
86

distinct public sources cited

See the registry

The gap

Today banks answer this with keyword lists, one card code, and self-declaration

$8.82T

Adjusted stablecoin transfer volume in the first half of 2026, more than all of 2024. Every mint, redemption, and ramp settles over bank rails. Source

$150

The average threshold at which traditional financial institutions alert on crypto exposure. Banks already do this work, with tools that were not built for it. Source

25 to 35

Sponsor banks under consent orders in 2024 and 2025. A 2026 OCC order faults a sponsor bank for not understanding the transactions in its payment processing line. Source

Why now

Regulators opened the door and kept the obligations

Banks may now serve crypto businesses, and stablecoin issuers are becoming BSA financial institutions. The obligation to know who is on the other side of a payment did not change. Every entry links to its source.

  1. March 2024

    Mastercard shuts down CipherTrace Armada, a product that identified crypto businesses in bank fiat data. Source

  2. March to May 2025

    OCC Interpretive Letters 1183 and 1184, FDIC FIL-7-2025, and Federal Reserve rescissions let banks take on crypto activities without prior approval. BSA obligations stay in force. Source

  3. July 2025

    The GENIUS Act is signed. Payment stablecoin issuers become Bank Secrecy Act financial institutions with full AML program obligations. Source

  4. August 2025

    The Fair Banking executive order prohibits debanking lawful businesses. Banks have to identify and risk-rate crypto customers instead of exiting them. Source

  5. January 2026

    Basel SCO60 capital treatment and DIS55 disclosure templates take effect. Banks under the standard must measure and publish their crypto exposure. Source

  6. February 2026

    New OCC community bank BSA/AML examination procedures take effect. Risk assessments must reflect the customers a bank actually has. Source

  7. April 2026

    FinCEN and OFAC propose AML rules for stablecoin issuers that count advanced monitoring technology as a mitigating factor in enforcement. Source

  8. June 2026

    OCC Bulletin 2026-24 proposes weekly and quarterly reporting forms for payment stablecoin issuers: PS-01 for activity and reserves, PS-02 for condition and income. Source

  9. July 2026

    Federal regulators miss the GENIUS Act's July 18 deadline for final rules. The Act takes effect on January 18, 2027, or 120 days after final rules, whichever comes first. Source

  10. September 2026

    Nasdaq Verafin and Stablecore partner to bring digital asset transaction data into Verafin. Crypto signals reach bank monitoring through integration partners, not from the bank's own fiat data. Source

  11. September 2026

    The Federal Reserve proposes reserve and capital rules for the stablecoin issuers it supervises, and an application process for banks that want to issue. Source

The product

What you get

One registry behind three layers: the assessment, the API and monitoring feed, and agent tools.

Exposure report

The sample file run through the live assessment: 25 transactions, classified and totalled by verdict.

Crypto counterparty

13

$5,944,770.00

Needs review

7

$17,190.50

No registry match

5

$860.92

Run the sample

Evidence drawer

Every match carries the record it matched and where that record came from.

input
FORIS DAX INC · wire
observable
Foris DAX, Inc.
type
legal_name
strategy
exact
source
crypto.com
observed
2026-08-08
review
reviewed

Precision

Ordinary businesses that share a word with a crypto brand are held for an analyst, not flagged.

  • KRAKEN SEAFOOD COMPANYReview

    containment on "Kraken" · 0.45 < 0.92

  • KRAKEN BREWING CO PORTLANDReview

    containment on "Kraken" · 0.45 < 0.92

API

Live on this site. No key on the early access tier.

POST /api/v1/classify
{
  "verdict": "crypto",
  "confidence": 0.978432,
  "strategy": "exact",
  "entity_name": "Bitstop",
  "categories": ["kiosk_operator"],
  "evidence": {
    "observable_value": "ATM Ops Inc",
    "review_status": "reviewed",
  ...

MCP server

Eight tools for AML agents, on the same engine.

  • classify_descriptor
  • classify_batch
  • registry_summary
  • lookup_entity
  • watch_counterparties
  • exposure_events
  • registry_changes
  • monitor_summary
What each tool does

Delivery

Into the monitoring stack you already run. Built means implemented and tested in the classifier service, which we set up for design partners.

  • Unit21 alertsBuilt

    Verified against Unit21's published API reference, pinned by contract tests.

  • WebhooksBuilt

    HMAC-SHA256 signatures, every attempt recorded.

  • CSV, JSON, markdownLive

    Ledger and report exports from the assessment and the API.

  • NICE ActimizeDraft

    Waiting for a partner's specification.

How it works

A registry of crypto businesses, mapped to how they appear in payment text

  1. 1

    The registry

    Exchanges, kiosk operators, brokers, OTC desks, payment processors, and stablecoin issuers, each mapped to the names that appear in bank data: legal names, trading names, and aliases. Not wallet addresses.

  2. 2

    The matching

    The payment text is parsed with the grammar for its rail: wire tags, ACH markers, or card descriptors. The counterparty fields are matched four ways: exact, without corporate suffixes, contained in longer text, and fuzzy. The best match gets a calibrated confidence.

  3. 3

    The evidence

    Every match shows the name it matched, the source of that name, the date it was checked, its review status, and the calibration bin behind the confidence. A bank has to be able to defend an attribution, so each one carries its evidence.

Provenance

The legal name rarely matches the brand

A wire carries the legal entity that holds the bank account, not the app name.

A keyword list of exchange brands never sees these names.

The same list flags a seafood company that happens to be called Kraken.

The registry maps each legal name to its brand, with a public source.

  • ATM OPS INC

    belongs toBitstopmsb.fincen.gov

  • RED LEAF CHICAGO, LLC

    belongs toDigitalMintmsb.fincen.gov

  • ELECTRIC SOLIDUS, INC.

    belongs toSwan Bitcoinswanbitcoin.com

  • FORIS DAX, INC.

    belongs toCrypto.com (US)crypto.com

  • PAYWARD, INC.

    belongs toKrakenkraken.com

  • PLUTUS FINANCIAL INC.

    belongs toAbradfi.wa.gov

Each line is a legal_name observable in registry v0.4.1, with its source host.

  1. Rule 1

    Mixed businesses are refused

    Stripe, JPMorgan, and SoFi Bank carry huge ordinary volume. Only a crypto subsidiary, such as PayPal Digital, Inc., is registered.

  2. Rule 2

    Ambiguous single words are capped

    Block, Strike, Abra, River, and Choice sit below the auto-classify threshold. They go to an analyst.

  3. Rule 3

    Partner banks are never matched

    Uphold's bank and Coinme's bank are stored as context for an analyst, never as names to match.

The rules, with the registry's own values

Who it is for

Teams that hold the exposure

Sponsor and BaaS banks

Consent orders cite banks for not understanding the businesses behind their payment flows. Counterparty classification is the control examiners describe.

Banks entering digital assets

Custody, trading, and stablecoin programs start with one supervisory question: what crypto exposure already moves over your rails?

Stablecoin issuers

The GENIUS Act made issuers BSA financial institutions. Mint and redeem legs are fiat wires, and they need bank-grade monitoring.

Community bank BSA teams

An exposure assessment you can defend, without a six-figure analytics contract or a consulting engagement.

Status

Where this stands

Live today

  1. The self-serve exposure assessment, on registry v0.4.1 with 127 entities.
  2. The early access API: classify, batch, assessment, calibration, and registry summary.
  3. The registry coverage page, with every count computed from the shipped data.
  4. Built and tested in the classifier service: monitoring, the registry change feed, signed webhooks, the Unit21 adapter, and the MCP server.

Next

  1. A design partner labels real records, and we refit the calibration on bank data.
  2. The Actimize mapping, from a partner's specification.
  3. Matching on ACH company identifiers and descriptor patterns.
  4. A retry queue for webhook delivery.
  5. More registry entries from the FinCEN review queue.

There are no customers yet, so none are named, and no metrics are invented. Market facts link to their sources. Confidence is calibrated on synthetic descriptors (table v0.4.1+2026-10-10), and every report and API response says so.

Get in touch

Two ways to start

Run an assessment on your own data

Send a period of wire, ACH, and card settlement records. You get the classified ledger and the report, with the evidence for every match and a list of what we could not identify.

Become a design partner

Today the confidence numbers are fitted on synthetic descriptors, because no bank has labeled a set with us yet. A partner who corrects our verdicts on real records makes the classifier measurably better and gets that accuracy first.

Run it on your own data.

Upload a CSV of wire, ACH, or card records. It is classified in memory and never stored.

Run a live assessmentEmail us