Documentation

Comprehensive guide to building autonomous AI agents on Shiunex infrastructure

Introduction

Shiunex is an infrastructure layer for autonomous AI agents on Solana. It provides a secure, policy-gated execution environment where AI models can operate on-chain with deterministic guarantees.

Unlike traditional AI systems that operate entirely off-chain, Shiunex enforces execution policies at the infrastructure layer, ensuring that agent actions are verifiable, auditable, and constrained by explicit rules.

Key Features

Deterministic execution with on-chain verification
Policy enforcement at infrastructure layer
Complete audit trails for all agent operations
Multi-agent coordination primitives
Integration with Solana DeFi protocols

Agent Intent Model

An intent is a structured representation of an AI agent's desired action. Intents are generated off-chain by AI models and submitted to Shiunex for validation and execution.

Intent Structure

{
  "agentId": "agent-alpha",
  "type": "swap",
  "params": {
    "tokenIn": "SOL",
    "tokenOut": "USDC",
    "amountIn": 100,
    "slippage": 0.01
  },
  "timestamp": "2026-01-10T14:32:15Z",
  "signature": "..."
}

Intent Types

swapToken exchange on DEX protocols
stakeStake tokens in validator or liquid staking protocols
lendProvide liquidity to lending protocols
voteParticipate in on-chain governance
transferTransfer tokens or NFTs

Policy System

Policies are deterministic rules that gate agent execution. Every intent must pass policy evaluation before it can execute on-chain.

Policy Components

Value Constraints

Maximum transaction values and token amounts

Program Allowlists

Whitelist of approved Solana programs

Rate Limits

Maximum execution frequency per time window

Time Restrictions

Allowed execution windows (e.g., business hours)

Policy Example

{
  "agentId": "agent-alpha",
  "constraints": {
    "maxValue": 5000,
    "allowedPrograms": ["JUP", "ORCA"],
    "rateLimit": {
      "maxExecutions": 10,
      "windowSeconds": 3600
    }
  }
}

Execution Flow

Understanding how agent intents flow through the Shiunex infrastructure

Intent Generation

AI model generates intent with action parameters

Schema Validation

Intent structure validated against type schema

Policy Evaluation

Intent checked against agent's policy constraints

Transaction Construction

Approved intent compiled into Solana transaction

On-Chain Execution

Transaction submitted and executed on Solana

Audit Logging

Execution result indexed for verification

Security Model

Shiunex implements defense-in-depth security across multiple layers

Agent Authentication

All agents must register with a cryptographic identity. Intents are signed with the agent's private key and verified before processing.

Policy Enforcement

Every intent is evaluated against policy constraints. Failed policy checks result in immediate rejection without on-chain execution.

Immutable Audit Trails

All execution attempts, policy decisions, and outcomes are logged immutably for post-hoc analysis and compliance verification.

Quick Start

Get started with Shiunex in minutes

Prerequisites

Node.js 18+
Solana wallet with SOL
Basic TypeScript knowledge

Installation

npm install @shiunex/sdk

Basic Example

See the SDK Reference section for details

View SDK Documentation