On-Chain Analysis Platform — 9-Chain Blockchain Analytics SaaS
On-chain analysis is the discipline of extracting fund-flow, behavioral, and risk signals directly from public blockchain transaction data. ChainAnalyzer is a Japan-built end-to-end on-chain analysis platform (refinancier, inc.) covering 9 blockchains in a single API and SaaS, with Neo4j graph analysis, a 3-model ML ensemble, 76+ AML detection rules, and AI-agent integration.
What on-chain analysis actually involves
Public blockchains (Bitcoin, Ethereum, etc.) record every transaction permanently and openly. On-chain analysis turns that raw stream into useful artifacts:
- Fund-flow tracing — graph traversal upstream / downstream, surfacing related addresses
- Risk scoring — sanctions, mixer, drainer, scam involvement classification
- Behavioral analysis — whale concentration, hold/sell trends, liquidity changes, arbitrage patterns
- Entity clustering — grouping multiple addresses controlled by the same operator
- Anomaly detection — ML to flag transactions outside the baseline distribution
- Compliance evidence — collecting audit trails for AML / KYT / FATF Travel Rule / SAR filings
Core technical stack of an on-chain analysis platform
- RPC / indexers — Etherscan, Alchemy, Helius, Blockstream Esplora pull transaction data
- Graph database — Neo4j or equivalent stores "address → tx → address" relationships for efficient traversal
- Heuristic detectors — rule-based identification of known illicit patterns
- Machine learning — unsupervised anomaly detection (Isolation Forest, AutoEncoder) plus graph ML (GraphSAGE)
- OSINT integration — Reddit, X, GitHub, ScamDB enrich the on-chain signal with off-chain context
- Reporting — PDF / JSON / Webhook output for internal teams, partners, or regulators
The ChainAnalyzer engine
Neo4j graph expansion (Follow Mode)
Watched addresses are expanded 1-3 hops via BFS, and related suspicious addresses are surfaced automatically — the direct functional analogue of Chainalysis Reactor's pivot. Follow Mode details.
3-model ML ensemble (anomaly scoring)
Rules miss novel patterns. We combine three ML models into a 0-1 anomaly score:
- Isolation Forest — statistical outlier detection
- AutoEncoder — reconstruction error as anomaly signal
- GraphSAGE — Neo4j subgraph topology embedding
See ML scoring writeup.
76+ heuristic detection rules
Chain-specific rule sets for EVM, Solana, and Bitcoin: address poisoning (we discovered a $5.3M operating ring), peel chain, drainer, rug-pull, and more. All 76+ detection rules.
Cross-chain bridge tracing
Six bridges detected via three methods (event log, calldata, memo): Wormhole, LayerZero, Stargate, deBridge, Across, NEAR Intents. Bridge investigation feature.
Real-world usage
- Crypto exchanges — deposit / withdrawal AML screening, suspicious activity reporting
- Institutional investors — counterparty risk assessment (OTC, DeFi)
- Audit firms / compliance consultants — AML program audits, PDF deliverables
- Law enforcement — seized-asset tracing, criminal-network visualization
- Security researchers — drainer / scam network analysis, vulnerability reports
- AI agents — autonomous counterparty screening via x402 (no subscription)
How ChainAnalyzer compares to other on-chain analysis platforms
The on-chain analysis space includes Chainalysis (Reactor / KYT), Elliptic, TRM Labs, Crystal Blockchain, Nansen, Arkham, Dune Analytics, and others. Specializations:
- AML / compliance focus: Chainalysis, Elliptic, TRM Labs, ChainAnalyzer
- Market & whale analytics: Nansen, Arkham, Glassnode
- Open queryable data: Dune Analytics, Bitquery
- Personal portfolio: Zerion, DeBank
For AML specifically, see Chainalysis alternatives — selection criteria.
Integration paths
- Web SaaS — instant browser use, Free tier with 10 scans/mo
- REST API — system integration via
tfk_ API key (Pro+) - MCP server — direct call from Claude, Cursor, etc. (Glama AAA score)
- x402 — AI agents pay per call ($0.008 USDC) without a subscription
30-second trial
Run a single on-chain analysis from your browser via the x402 demo with Phantom on Base — $0.008 USDC, no signup.
Related pages