GUARDRA Protocol

Overview

GUARDRA is a risk and insurance layer for autonomous AI agents on Arc, with coverage and payouts denominated in USDC. The protocol is in development; this documentation describes the intended design.

What is Guardra?

A protection layer that sits between an agent's intent and its on-chain execution, pricing risk per action and settling defined failures.

Agent → Risk Engine → Policy → Execution → Verification → Settlement

Agent Insurance

Coverage attaches to a specific autonomous action rather than an open-ended account. Terms are programmatic: premium, limit, conditions and verification rules.

Risk Engine

Scores agents and actions from history, reputation, collateral, execution reliability, dispute history and behavior consistency.

// illustrative only
const quote = await guardra.risk.score({
  agent: "0xAGENT",
  action: { type: "swap", notional: 5_000, asset: "USDC" },
});
// → { score: 72, tier: "MEDIUM", premiumBps: 140 }

Coverage

Planned coverage classes: transaction, execution, agent failure, protocol risk and data/API risk. None are live.

Insurance Pools

USDC liquidity providers underwrite coverage, receive premiums and absorb approved claims according to pool parameters.

Claims

SUBMIT → VERIFY → ASSESS → APPROVE → SETTLE → PAYOUT

Agent Reputation

A persistent agent passport links wallet, operator and performance history so risk can be attributed and priced over time.

Security

Contracts will be audited before deployment. Until then, no funds are accepted and no contract address is published.

Architecture

AI AGENTS → AGENT IDENTITY → RISK ENGINE
  → { COVERAGE, POLICY } → ON-CHAIN ACTION → VERIFICATION
  → SUCCESS → SETTLEMENT
  → FAILURE → CLAIM → USDC PAYOUT

Roadmap

FAQ