Crypto compliance
intelligence

Screen any wallet for sanctions, mixer and laundering risk — with a risk score, the regulatory licenses behind it, and the known illicit wallets it behaves like. A live API a bank’s compliance team can call today.

Get Started
Explore
835M / 15.8Bnodes / edges · joint 5-chain EVM graph
1.5BBitcoin addresses scored
0.83recall @ 0.1% FPR · illicit detection (ETH)
1,256VASP entities · 9 jurisdictions
6 chainslive · ETH Tron BTC Base Arbitrum Gnosis

Capabilities

01

Wallet Screening & Risk Scoring

One API call per address returns sanctions exposure, entity category and a risk tier — live in production across six chains. Sanctions verdicts run against the live OFAC SDN list, refreshed within the hour and covering every chain OFAC publishes addresses for — 16 chains, 989 SDN addresses today, Bitcoin and Monero included. The graph-neural score is trained on the full 835M-address, 15.8B-edge five-chain EVM graph with per-chain normalisation: recall 0.83 on ETH and 0.58 on Tron at a 0.1% population false-positive rate. Base, Arbitrum and Gnosis are covered by zero-label transfer — 0.86 / 0.82 / 0.997 with heads that never saw those chains. Bitcoin is scored by a dedicated feature pipeline over 1.5B addresses. Explainable, not a black box.

02

AI-Agent Wallet Detection

Identify autonomous AI-agent wallets from behavioural patterns — timing, amounts, scheduling — at AUROC 0.9987. The emerging on-chain actor incumbents don't score.

03

Network Reconstruction

Build graphs from heterogeneous signals — transactions, registries, blockchain, media. Direct feeds from proprietary Bitcoin and Ethereum full nodes. Automated entity resolution across data silos.

04

Hidden Structure Detection

Identify hubs, communities, and anomalies with GNN and network science. Structural Asymmetry Index reveals power concentration invisible to standard analysis.

05

Scenario Simulation

Agent-based modelling for behavioural adaptation and intervention outcomes. Stress tests and sanction impact — before decisions are made.

06

Decision-Ready Output

Intelligence briefs, structural diagnostics, scenario comparisons. Actionable — fragility, leverage points, intervention effects.

Live screening — examples

One address in. A plain-language verdict out.

Sanctions, mixers, laundering paths and AI-agent wallets — one screening line, six chains live. Six real Ethereum results below.

Risk tier1 = regulated & safe→5 = sanctioned or mixer, do not transact

0x098b…2f96Ethereum
Sanctioned
Tier 5 of 5
Critical — block

On the OFAC-SDN sanctions list. Transacting with it is illegal.

Sanctions hitHighest risk tier
JSON — for technical specialists
{
  "address": "0x098b716b8aaf21512996dc57eb0615e2383e2f96",
  "chain": "ethereum",
  "sanctions_hit": true,
  "category": "personal",
  "risk_tier": 5,
  "flags": ["sanctions_hit"],
  "model_version": "60.amro-s1"
}
0x0389…6003Ethereum
Mixer
Tier 5 of 5
Critical — block

A Tornado Cash address — used to obscure where funds came from.

Mixer — directFunds obfuscation
JSON — for technical specialists
{
  "address": "0x03893a7c7463ae47d46bc7f091665f1893656003",
  "chain": "ethereum",
  "category": "mixer",
  "risk_tier": 5,
  "mixer_exposure": { "direct": true, "mixer_names": ["tornado_cash"] },
  "flags": ["mixer_direct"],
  "model_version": "60.amro-s1"
}
0x1f90…c326Ethereum
AI-agent · 84%
Tier 4 of 5
Elevated — review

Run by software, not a person — periodic, clustered, MEV-driven activity.

Perfect periodicityAmount clusteringMEV activity
JSON — for technical specialists
{
  "address": "0x1f9090aae28b8a3dceadf281b0f12828e676c326",
  "chain": "ethereum",
  "category": "ai_agent",
  "agent_probability": 0.84,
  "risk_tier": 4,
  "indicators": [
    { "indicator": "perfect_periodicity", "score": 0.90 },
    { "indicator": "amount_clustering",  "score": 0.80 },
    { "indicator": "mev_activity",        "score": 0.71 }
  ]
}
0xd8da…6045Ethereum
Clean
Tier 3 of 5
No adverse signal

No sanctions, no mixer, no agent signal — safe to proceed.

No sanctions exposureNo mixer exposure
JSON — for technical specialists
{
  "address": "0xd8da6bf26964af9d7eed9e03e53415d37aa96045",
  "chain": "ethereum",
  "sanctions_hit": false,
  "category": "unknown",
  "risk_tier": 3,
  "flags": [],
  "model_version": "60.amro-s1"
}
0xdc76…00d3Ethereum
Exchange wallet
Tier 2 of 5
Low risk — known exchange

A known exchange deposit address. Counterparty is an exchange, not an anonymous wallet — low risk.

Exchange-attributedNo sanctions exposure
JSON — for technical specialists
{
  "address": "0xdc76cd25977e0a5ae17155770273ad58648900d3",
  "chain": "ethereum",
  "category": "exchange_depositor",
  "risk_tier": 2,
  "sanctions_hit": false,
  "flags": [],
  "model_version": "60.amro-s1"
}
0x7166…75d3Ethereum
Licensed VASP
Tier 1 of 5
Regulated — lowest risk

Coinbase — a licensed exchange, regulated across six jurisdictions (FCA, FinCEN, MAS, NY DFS, CSSF, JP FSA).

Licensed VASP6 active licenses
JSON — for technical specialists
{
  "address": "0x71660c4005ba85c37ccec55d0c4493e66fe775d3",
  "chain": "ethereum",
  "category": "vasp",
  "risk_tier": 1,
  "vasp": {
    "name": "Coinbase",
    "n_jurisdictions": 6,
    "licenses": [
      { "jurisdiction": "GB", "regulator": "FCA", "status": "active" },
      { "jurisdiction": "US", "regulator": "FinCEN", "status": "active" }
    ]
  },
  "sanctions_hit": false,
  "flags": []
}

Free Research tier: 25 screening checks a month, no card — API key in minutes.

Request an evaluation key

Shadow Networks Lab

See the
invisible

Hidden channels shape global flows but resist direct measurement. We map financial-technological networks, locate structural asymmetries, and model how systems reconfigure under constraints.

Structural diagnostics
Stress tests
Scenario comparisons
Intelligence briefs
Agent detection
Blockchain explorer
Click to interact
Hub
Entity
Shell
Mixer

Use Cases

Identify autonomous AI agents operating wallets in real time.

Behavioral analysis: detect bots by timing, amount patterns, and scheduling
Machine learning anomaly detection on transaction networks
Automated discovery of hidden money flow pathways between clusters
Interactive wallet relationship maps with risk scoring

Open by design

OPEN-CORE

Labelling toolchain, open

The tooling behind our 301,304-address label set, the adversarial simulator, the detector's training code, and the paper's evaluation harness — registry builders, VASP-register scrapers for 9 jurisdictions, TagPack generators, agent-based simulation, the GNN architecture with its warm-start recipe and threshold calibration, and the lead-time replay with its honest event-level null — is released as open source under Apache-2.0, with byte-reproducible pipelines verified in CI.

openlabels + openabm + openheads + openeval — live now
github.com/ai-decisions/openlabels →github.com/ai-decisions/openabm →github.com/ai-decisions/openheads →github.com/ai-decisions/openeval →
PAPER

Methods on the record

We publish FPR, recall, sample sizes and label provenance — including the public Elliptic benchmark, where our model scores AUC 0.939 against the published 0.796 baseline. The methodology preprint is on SSRN (DOI 10.2139/ssrn.7366202).

Preprint on SSRN · Reproducible · DD-defensibleRead the preprint →
FREE TIER

Research corpus, free

The Research tier is free forever: 25 screening checks a month plus the full intelligence corpus — 10,500+ documents, the entity graph with GEXF/GraphML export, intelligence digests and weekly reviews. No credit card required.

Free forever · no cardGet a free key →
FREE CHECKER

Check a wallet, free

A free, rate-limited wallet check against the same live screening line the API serves — sanctions, mixer exposure, risk tier. No account required.

Live nowCheck a wallet →

See what’s hidden

We respond when a message contains a clear question and enough context to scope. One domain, no exceptions.

Get Started

Defensive, regulatory, and analytical purposes only