AI-Native Non-Custodial Wallet
An AI-powered crypto wallet that explains transactions, not controls them. Deterministic • Non-custodial • Audit-ready • Research-grade
🌍 Overview
This project is a next-generation mobile crypto wallet inspired by the architectural principles behind Tether QVAC and modern secure wallet research.
The wallet integrates AI explainability, deterministic security, and zero-knowledge verification while strictly preserving the core crypto principle:
Users control keys. AI never does.
✨ Key Features
🔐 Non-Custodial by Design
Private keys never leave the device
Secure enclave / keystore support
No AI access to signing or broadcasting
🧠 AI-Powered Transaction Explainability
Deterministic transaction explanations (baseline)
Domain-aware explainers:
NFT (mint, transfer, approval)
DeFi (swap, deposit, borrow)
Bridge (cross-chain, lock, delay)
Optional offline sandboxed LLM for extended explanations
🌍 Multilingual (ID-First)
Bahasa Indonesia as default language
English fallback
Easily extensible to other languages
🔬 ZK-Proofed Inference (Research-Grade)
Zero-Knowledge proofs that explanations were generated correctly
No leakage of private keys or sensitive data
Verifiable by auditors, DAOs, or regulators
🧱 Audit-Ready Architecture
Deterministic logic first
AI is optional and sandboxed
Kill-switches at every AI layer
🧩 Architecture Overview
User ↓ Transaction Builder ↓ Deterministic Explainer (MANDATORY) ↓ Domain Explainer (NFT / DeFi / Bridge) ↓ LLM Sandbox (OPTIONAL, OFFLINE) ↓ ZK Proof Generator (OPTIONAL) ↓ User Confirmation → Signing
📁 Repository Structure
apps/ └─ mobile/ # React Native wallet app core/ ├─ wallet-core/ # Key management & signing ├─ contract-explainer/ # Deterministic explainers │ ├─ domains/ # NFT / DeFi / Bridge logic │ └─ i18n/ # ID-first multilingual support ├─ ai-llm/ # Offline sandboxed LLM └─ zk-proof/ # Zero-knowledge inference proofs security/ ├─ SECURITY.md └─ THREAT_MODEL.md
🔐 Security Principles
❌ No AI signing
❌ No AI transaction mutation
❌ No network access for LLM
✅ Deterministic explanations first
✅ Explicit user confirmation required
For details, see SECURITY.md.
🧠 AI Philosophy (QVAC-Style)
Principle Description
Explain AI explains what happens Not Decide AI never decides or executes Deterministic First AI augments, never replaces Kill-Switch AI can be disabled safely
🧪 Example User Experience
NFT Approval
⚠️ Anda memberikan kontrol penuh atas SEMUA NFT Anda ke alamat lain.
DeFi Borrow
📉 Anda meminjam aset. Risiko likuidasi berlaku.
Bridge Transfer
🌉 Anda memindahkan aset lintas blockchain. Proses dapat tertunda atau gagal.
🔬 Research & Compliance Ready
This project is suitable for:
Core infrastructure proposals
Security audits
Academic research (AI + ZK + crypto)
Regulated or semi-regulated environments
🚀 Roadmap Summary
✅ Phase 1: Wallet Core + Deterministic Explainer
✅ Phase 2: Domain Explainers + Multilingual
✅ Phase 3: Offline LLM Sandbox
✅ Phase 4: ZK-Proofed Inference
🔵 Phase 5: Governance-signed models & proofs
See ROADMAP.md for full details.
This software:
Does NOT provide financial advice
Does NOT guarantee transaction safety
Is a tool for informed consent, not automation
Users are fully responsible for their actions.
🏁 Status
🟢 Architecture complete 🟢 Security model defined 🟡 Production hardening ongoing 🔵 External audit pending
🤝 Contributing
This repository welcomes:
Security researchers
Wallet engineers
Cryptography & ZK researchers
All contributions must respect the non-custodial & deterministic-first principle.
📜 License
Apache-2.0 (recommended for infrastructure & research projects)
Built for users who deserve to understand before they sign.