Concepts

The core building blocks: identity, smart accounts, policies, permissions, and enforcement.

Agents

Each agent is backed by an AgentSmartAccount contract (ERC-4337) and an identity record in IdentityRegistry. Generate a dedicated keypair so your bot has its own wallet, completely separate from yours.

Generate bot keypair — never share your personal keys
On-chain enforcement — disallowed transactions revert in validateUserOp
Connected wallet stays the owner — full withdrawal rights at all times

Policies

Policies define the rules for what actions an agent can perform. They specify allowed actions, assets, constraints, and time-based validity.

Action allowlists (swap, transfer, etc.)
Value and volume constraints
Time-based validity windows

Permissions

Permissions link agents to policies. When you grant a permission, you're authorizing an agent to act according to a specific policy's rules.

Connect agents to policies
Set expiration dates
Mint on-chain for enforcement

x402 Payments

x402 payments use stablecoin authorizations (EIP-3009 transferWithAuthorization) that settle off-account via the account's EIP-1271 signature. Aegis routes them through authorizeX402Payment so they cannot bypass policy.

Integration note: the exact payment digest must be pre-authorized on-chain (authorizeX402Payment) before the facilitator verifies it. The bot signs a UserOp that the Aegis backend relays (gas-sponsored) — the owner key is never needed at runtime. Skipping this returns an empty 402. See the smart-accounts skill for the full flow.

Policy-checked before the payment can settle
Counts toward daily volume and per-tx limits
Captured in the audit log with on-chain proof

Audit Logs

Every action in the system is logged for compliance and debugging. View validation requests, policy changes, and permission grants.

Complete action history
Filter by event type
Export to JSON/CSV

Example: a trading agent policy

A policy that lets an agent swap USDC for WETH on Uniswap within strict limits

policy.json

{
  "name": "ETH Trader",
  "description": "Autonomous agent that swaps USDC for WETH on Uniswap",
  "enforceOnChain": true,
  "policies": {
    "allowedActions": ["swap", "transfer"],
    "allowedProtocols": ["uniswap-v3"],
    "assetLimits": [
      { "asset": "0x8335...USDC", "maxPerTx": "100000000", "maxDaily": "1000000000" },
      { "asset": "native", "maxPerTx": "1000000000000000000", "maxDaily": "5000000000000000000" }
    ],
    "timeWindow": {
      "validFrom": "2024-03-01T00:00:00Z",
      "validUntil": "2024-03-31T23:59:59Z"
    }
  }
}

If this agent tries to execute a 2.51 ETH swap, the smart contract rejects it before it reaches Uniswap. A compromised agent still cannot exceed the limits — they are enforced by the chain, not by the API.

How Aegis compares to other approaches

Where Aegis fits in the landscape of agent safety mechanisms

Capability	Aegis	Multi-Sig	Rate Limiting	Off-Chain Validator
On-chain enforcement	Yes	No	No	No
No private key sharing	Yes	Partial	Yes	Yes
Detailed action policies	Yes	No	Partial	Yes
Cryptographic proof of enforcement	Yes	Partial	No	No
Multi-agent management	Yes	Yes	Partial	Partial
Verified, auditable contracts	Yes	Partial	Partial	No