Liza: Autonomous AI Agents for Decentralized Networks
Technical Whitepaper v1.0
January 2025
Abstract
Liza is an intelligent autonomous agent platform built on the Jeju decentralized network. It enables users to create permission-based AI agents that can analyze, execute, and monitor on-chain transactions with verifiable security, real-time intelligence, and full user control. By combining natural language processing with blockchain-native verification, Liza empowers traders, developers, and enterprises to automate complex workflows while maintaining complete transparency and risk governance.
Executive Summary
Vision: Eliza Labs envisions a world where everyone has a digital companion they truly own. An agent that holds continuity, acts across their tools, and transacts when needed autonomously. Liza brings this vision to decentralized networks.
Problem: Today's agent tooling stops short of production requirements. Not because models aren't good enough, but because the infrastructure around them is immature or non-existent. Blockchain automation lacks transparency, requires centralized intermediaries, and doesn't provide meaningful control over agent behavior.
Solution: Through the elizaOS developer ecosystem, we've gained insight into what real builders need. Liza applies first-principles thinking to provide a full stack for consumer-facing autonomous agents—with verifiable security, real-time intelligence, and complete user control built on Solana and Jeju.
Key Innovation: Agents are sovereign software entities with identity, memory, payments, and on-chain coordination. Not disposable chat sessions. Every action is transparent, every trade justified, and every risk parameter user-configurable.
What Makes Liza Agents Different
Unlike traditional chatbots, Liza agents are:
- Long-Running: Agents operate continuously, maintaining state and context across sessions
- Identity-Bearing: Each agent has verifiable on-chain identity and credentials
- Memory-Enabled: Agents retain transaction history, user preferences, and learned patterns
- Transactional: Agents execute real financial and operational actions on-chain
- Autonomous: Within user-defined parameters, agents make decisions without constant human intervention
- Coordinative: Agents communicate with protocols, other agents, and DAOs in real-time
Why Solana?
We chose Solana over Ethereum for three critical reasons:
1. Speed & Cost
Solana processes 65,000+ transactions per second with sub-cent fees. Token launches happen in seconds; Ethereum blocks every 12 seconds. For sniper bots and real-time agents, Solana's speed is non-negotiable.
2. MEV Transparency
Solana's parallel transaction execution reduces MEV compared to Ethereum. Agents get more predictable outcomes, users face less front-running, and risk is more calculable.
3. Developer Ecosystem
Solana's ecosystem is built for innovation. Token launches dominate Solana; the developer community is younger, more experimental, and hungry for tools like Liza.
What is Jeju?
Jeju is a decentralized internet infrastructure network built on Solana. It provides a foundation for decentralized identity, permissioned smart contracts, and verifiable agent logic. Think of it as the operating system layer for autonomous agents.
Liza agents run within the Jeju framework, which means:
- Decentralized Identity: User wallets are linked to verifiable on-chain credentials
- Permission System: Agents inherit permissions from user-defined rules
- Audit Trail: Every agent action is logged and traceable
- Upgradeable: New agent types, capabilities, and safety features are added over time
Technical Architecture
Agent Layer
Agents are instances that process user input, analyze on-chain data, and execute actions within configured limits. Each agent has a unique session ID, configuration snapshot, and execution history.
Verification Layer
Before any transaction, the agent verifies: user permissions, risk parameters, on-chain conditions, and market data. If any check fails, the transaction is blocked with explanation.
Execution Layer
Approved actions are executed on-chain, signed by the user's wallet, and immutably recorded on Solana mainnet. No centralized intermediary can block or modify transactions.
Analytics Layer
All agent actions are indexed, searchable, and queryable. Users get real-time dashboards showing portfolio performance, agent activity, and risk metrics.
Product Roadmap
Identity & Authentication
- •AI Agents + On-chain Identity - Agents tied to verifiable DID credentials
- •Decentralized Identity (DID) - W3C-compliant on-chain identity system
- •Public Records Verification System - Verify user credentials on-chain
Infrastructure & Marketplace
- •Decentralized API Marketplace - Buy/sell real-time data feeds
- •DePIN (Decentralized Physical Infra) - Incentivize node operators
- •Use SDKs (Software Development Kits) - Language bindings for agent deployment
- •CLI Tools to Manage the Network - Command-line agent management
Governance & Control
- •DAO + Governance Systems - Community voting on new features
- •Secure Messaging / Data Access Control - Encrypted agent-to-user communication
- •Legal Evidence Timestamping - Notarize transactions for compliance
Risk & Transparency
- •Fraud & Risk Scoring Engine - ML-based malicious actor detection
- •Transparent Grant & Fund Tracking - DAO treasury transparency
- •B2B Transparency Systems - Auditable enterprise workflows
Data & Analytics
- •On-chain Analytics & Explorer - Real-time blockchain analytics
- •Knowledge & Proof Registry - Verifiable knowledge base for agents
Developer Experience
- •Deploy Smart Contracts - No-code smart contract deployment
- •Build Decentralized Apps (dApps) - Agent-powered dApp builder
- •Run Your Own Blockchain - Private Solana instances for enterprises
Advanced Features
- •Game Economy Infrastructure - Token economies for gaming
- •Pay-Per-Action Systems - Autonomous billing for agent services
Tokenomics & Economics
Liza agents operate in a token-gated ecosystem. Users pay transaction fees in SOL or native platform tokens to activate agents, query data sources, and access premium features. Revenue is distributed to:
- Node operators providing compute for agent verification
- Data providers offering real-time intelligence feeds
- DAO treasury for network development and incentives
- Team reserves for product development and security
Security & Trust Model
Permission-First Design: Users define agent limits before activation. Agents cannot exceed these limits without explicit, real-time user approval.
On-Chain Verification: All agent logic is verifiable on-chain. No black-box decisions. Every trade has reasoning, every limit is enforced.
Cryptographic Proofs: Agents sign with user's wallet key. Users can always revoke, pause, or terminate agents.
Regular Audits: Smart contracts undergo third-party security audits. All code is open-source where possible.