Core Concepts

AUTONEX is built on three fundamental concepts that work together to enable safe autonomous agent operations: Intents, Policies, and Execution Receipts.

Intents

An intent is a high-level description of what an agent wants to do. Unlike raw transactions that specify exactly how to execute an operation, intents describe the desired outcome and let AUTONEX determine the best execution path.

Intent Structure

Every intent includes four key components:

• Type: The action category (swap, transfer, stake, etc.)
• Parameters: Action-specific configuration
• Constraints: Limits like slippage, deadline, or price bounds
• Context: Agent identity, timestamp, and execution preferences

{
  "type": "swap",
  "from": { "token": "SOL", "amount": 5 },
  "to": { "token": "USDC" },
  "maxSlippage": 0.02,
  "deadline": 1735200000
}

Supported Intent Types

Policies

Policies are on-chain rules that govern agent behavior. Every intent must pass all applicable policies before execution. Policies transform autonomous agents from uncontrolled systems into governed infrastructure.

Policy Categories

Financial Policies

Control spending, position sizes, and exposure limits:

• Transaction limits (per transaction, daily, monthly)
• Position size restrictions
• Portfolio concentration limits
• Minimum reserve requirements

Risk Policies

Protect against market risks and execution failures:

• Maximum slippage thresholds
• Price deviation limits
• Volatility-based circuit breakers
• Correlation exposure limits

Protocol Policies

Define which protocols and programs agents can interact with:

• Whitelist of approved programs
• CPI restrictions on cross-program calls
• Protocol-specific parameter constraints
• Account validation rules

Temporal Policies

Time-based restrictions and cooldown periods:

• Trading hours (e.g., only during market hours)
• Rate limiting between actions
• Cooldown periods after losses
• Time-locked permissions

Policy Example

{
  "name": "trading-limits",
  "rules": [
    {
      "type": "transaction_limit",
      "amount": 10,
      "token": "SOL",
      "period": "daily"
    },
    {
      "type": "slippage_protection",
      "maxSlippage": 0.02
    },
    {
      "type": "whitelist_programs",
      "programs": ["JUP...", "ORCA..."]
    }
  ]
}

Execution Receipts

Every execution generates an immutable receipt stored on-chain. Receipts provide a complete audit trail and enable reproducible verification of agent actions.

Receipt Contents

• Original Intent: The agent's requested action
• Policy Results: All evaluated rules and their outcomes
• Execution Details: Actual execution path and results
• Blockchain Data: Transaction signature, block, timestamp

{
  "signature": "4qVz...",
  "intent": { /* original intent */ },
  "policyResults": {
    "passed": true,
    "evaluations": [
      { "rule": "transaction_limit", "passed": true },
      { "rule": "slippage_protection", "passed": true }
    ]
  },
  "execution": {
    "status": "success",
    "actualSlippage": 0.015,
    "amountOut": 498.25
  },
  "timestamp": 1735199850
}

How They Work Together

The interaction between intents, policies, and receipts creates a secure execution flow:

This architecture ensures that agents can operate autonomously while remaining within predefined safety boundaries. The combination of intent-based abstraction, policy-gated execution, and immutable receipts enables production-grade autonomous systems.