About SolSentry

Operator threat intelligence for
Solana.

Solo-built. Free. Open-core. Live mainnet since April 2026 — tracking the wallets behind serial rug deployments before they catch the next buyer.

All numbers live. Click any stat to verify against the public API.

Why SolSentry exists

A single operator can rug, rename, and redeploy across many wallets and tokens. Retail traders see unrelated tokens; existing tools score each one in isolation — by the time they flag a pattern, the operator has already rotated to a fresh wallet.

The unit of analysis is wrong. The token is disposable. The operator is the persistent identity. SolSentry is the operator graph that didn’t exist — a cross-token, cross-wallet database of who deploys what on Solana, computed live from the chain, free for retail.

Started January 2025. First mainnet write April 8, 2026. A tracked operator wallet (4kxscute…) remains a live sample for public verification.

What makes it different

Unit of analysis
Others
One token at a time
SolSentry
Cross-token operator graph
Discovery latency
Others
Post-rug (after the drain)
SolSentry
Pre-rug (at deploy time)
Architecture
Others
Static rules engine
SolSentry
ALife agents that evolve
Audience
Others
Enterprise / English-only
SolSentry
Free retail + PT-BR consumer-first
Audit trail
Others
Black box / proprietary scores
SolSentry
Reproducible from public outcome_predictions.json
Cost
Others
Enterprise contracts
SolSentry
Free + per-call x402 (no subscription required)

How the brain works

Three-stage progressive scan (account info → DAS metadata + holder concentration → bot-cluster forensics + Token-2022 extension scoring) drives a multi-signal resolver with parallel resolution windows.

An autonomous brain layer — investigator, retract_engine, anomaly_seeker, plus stablecoin-flow and CEX-deposit tracers — runs investigation pipelines at roughly zero LLM tokens per call, replacing manual analysis.

A self-tuning MetaLearning system snapshots scanner DNA at every prediction, waits for outcome, and feeds accuracy back via auto_adjust() every 20 ticks. The system tunes itself.

Layer 1 · ✅ production
Known detectors
sham, drain, cluster, ring, privacy, first-funder — patterns pre-defined by humans
Layer 2 · 🟡 stub demo-ready
Anomaly memory
stores unclassified patterns · feature vector + wallet context · emits candidate when N similar accumulate
Layer 3 · 🔴 P0 next sprint
Research agent
autonomous detector breeder · LLM-generated detector code · backtest gate · admit if precision ≥70%

The honest numbers

Aggregate metrics on this page refresh from api.solsentry.app/v1/stats every 60 seconds.

  • Aggregate accuracy 79.3% across 220,216 predictions

Trust & transparency

Team

Crash Diniz — founder, sole developer. Self-taught since the early 2000s: Slackware, Unix, Oracle networking. No university, no bootcamp. Started learning Python in January 2025. Currently: 220,216+ mainnet predictions, full async architecture, and live precision metrics from the public API — solo, in Brazil.

Sena — the AI persona. Surfaces operator threat context in human language, in PT-BR or EN. Powered by Anthropic. Tone: senior security analyst, evidence-first, no sensationalism.

Acknowledgments: Mert Mumtaz (Helius CEO) for the open-source Haradrim patterns we credit in our graph viz. Pedro Marafiotti (Superteam BR · The Garage). Every contributor to the public MCP + docs repos.

What we'll never do

❌ Issue a token. Sem token. Ever.
❌ Multi-chain dilution — Solana depth is the moat
❌ Gate basic operator verdict behind paywall
❌ Sensationalize — every claim auditable from the API
❌ Sell user data — we don't collect it
❌ Replace RugCheck or competitors — we complement

Get involved