Every layer of protection, running on your terms.
SKEP3 combines EVM simulation, on-device AI, and a real-time threat database into a single, local-first security layer — with no cloud, no subscription, and no compromise.
EVM Transaction Simulation
See exactly what happens before you sign.
SKEP3 forks the Ethereum mainnet at the current block using Foundry's Anvil, then replays your pending transaction against this local state. The result is a cryptographically accurate preview of every token movement, every state change, and every approval — before a single gas unit is spent.
- Full EVM-faithful simulation via Anvil fork
- Exact token delta computation (ERC-20, ERC-721, ERC-1155)
- Catch approve-all and sweepETH before they execute
- Average simulation time under 200ms on mainnet fork
On-Device AI Threat Analysis
Neural-network threat scoring with zero cloud latency.
After simulation, SKEP3 feeds the structured output into a locally-running LLM (via Ollama). This model — trained on thousands of real drainer events, rug pulls, and honeypot interactions — produces a plain-English risk report in milliseconds, classifying threats from LOW to CRITICAL.
- Runs via Ollama — any compatible model supported
- Trained on real Web3 threat data and exploit patterns
- Structured JSON output minimises hallucinations
- 98% drainer detection rate, 2% false positive rate
Drainer & Malicious Contract DB
Block known threats instantly — zero simulation required.
SKEP3 maintains an on-device database of known malicious contract addresses, drainer EOAs, and flagged token contracts. Every transaction is first checked against this index before simulation — enabling sub-millisecond blocking of known bad actors without burning compute.
- 14,200+ known malicious addresses indexed
- Updated continuously from on-chain forensics and community reports
- Includes known drainers, rug pull contracts, and phishing deployers
- Offline-capable — full DB cached locally
Local-First Proxy Engine
Sits between your wallet and the chain. Invisibly.
SKEP3's proxy intercepts all outgoing JSON-RPC calls from your wallet. It reads eth_sendRawTransaction requests, extracts the transaction payload, routes it through simulation and AI analysis, then decides whether to forward, block, or warn. Your wallet needs zero configuration beyond a single RPC URL change.
- Drop-in replacement RPC — no wallet plugin required
- Compatible with MetaMask, Rabby, Phantom, and any JSON-RPC wallet
- WebSocket feed for live threat events at ws://localhost:9545/feed
- Full request/response logging for audit trails
Multi-Chain Support
Ethereum live. Solana coming.
SKEP3 currently provides full protection for Ethereum mainnet and any EVM-compatible chain (Arbitrum, Optimism, Base, Polygon). Solana support is on the roadmap for Q3 2026, extending the same simulation-first model to Solana Virtual Machine transactions.
- Ethereum mainnet — fully live and production-ready
- All EVM L2s supported: Arbitrum, Optimism, Base, Polygon
- Solana SVM simulation planned for Q3 2026
- Configurable RPC per chain via config file
Zero-Trust Security Model
Your keys never leave your machine. Ever.
SKEP3 never touches your private keys. The proxy intercepts RPC calls but does not hold wallet credentials. The AI model, EVM simulation, and threat database all run within your network boundary — nothing leaves.
- Private keys remain 100% client-side — never transmitted
- No external API dependencies or cloud callbacks
- Zero telemetry — no usage data sent externally
- Fully auditable and open-source
How SKEP3 compares
Against industry-standard cloud-based alternatives.
| Feature | SKEP3 | Blowfish | Pocket Universe |
|---|---|---|---|
| Runs on your infrastructure | |||
| Zero monthly cost | |||
| Privacy — no data sent to cloud | |||
| On-device AI analysis | |||
| EVM simulation (Foundry-grade) | |||
| Drainer signature database | |||
| Cannot be censored or taken down | |||
| Solana support (roadmap) |
What's under the hood
Production-grade tools, not homegrown glue.
Ready to deploy?
Get SKEP3 running locally in under 2 minutes.