Brain Protocol

The Brain Protocol is the open infrastructure layer that powers the agentic execution layer of the Brain platform. It gives AI agents a programmable financial identity, structured permissions, and the ability to pay for services autonomously, all on-chain.

The Brain Protocol is open and standards-based. Any developer can build on it directly using the Brain SDK, independently of the Brain platform.

---

What the Protocol Provides

🔐 Programmable Accounts

ERC-4337 smart accounts with policy-aware validation. Users retain custody; agents get scoped authority.

🪪 Agent Identity & Reputation

Persistent on-chain agent IDs and verifiable reputation via ERC-8004. Trust that is portable across platforms.

📋 Policy Enforcement

Structured policies define spend limits, allowed assets, time windows, and approval thresholds — enforced on-chain.

⚡ x402 Machine Payments

Agents pay for services over standard HTTP in real time — no API keys, no subscriptions, no manual billing.

---

How It Works

The protocol is implemented as a layered stack. Each layer is independent and validates separately, but they compose into a single deterministic execution pipeline.

1

Identity

Every agent receives a persistent agentId registered on-chain via the ERC-8004 agent registry. This ID is portable across applications and chains. Reputation accumulates from every verified action.

2

Permissions

Users attach structured policies to agents, defining spend limits, allowed contracts, permitted assets, time windows, and approval thresholds. Policies are hashed and enforced at the smart account level.

3

Execution

Agents propose actions. The policy engine evaluates each proposal against the agent's active policy and issues a signed approval proof. The proof is attached to a UserOperation submitted via ERC-4337.

4

Payments

When an agent requests a paid resource, the server responds with a standard HTTP 402. Brain parses the payment requirement, runs a policy check, settles on-chain, and retries the request automatically.

5

Auditability

Every action is linked to the agent's agentId, the active policy hash, and an on-chain transaction. Validation records are written to the ERC-8004 registry and reputation is updated accordingly.

---

Built on Open Standards

The Brain Protocol does not define new standards. Rather, it implements and composes existing ones into a cohesive system.

Standard	Role
ERC-4337	Smart account model and UserOperation-based execution
EIP-7702	Delegated execution for existing EOAs without address migration
ERC-8004	Agent identity, validation records, and portable reputation
ERC-8183	Conditional commerce — escrowed jobs and contingent payments
ERC-7902	Wallet capability discovery
ERC-7769	Standard bundler infrastructure compatibility
x402	HTTP-native machine-to-machine payments

Because Brain is built on open standards, external systems — wallets, applications, other agents — can interact with Brain agents without proprietary integrations or the Brain SDK.

---

Security Guarantees

The protocol enforces security across every layer.

No Unauthorized Execution

Every action requires valid SIWX authentication and a signed policy proof from the Brain policy engine. The PolicyValidator contract is invoked inside validateUserOp — any UserOperation without a valid proof is rejected at the smart account level before any funds move.

Full Auditability

All significant actions are linked to the agent's agentId, the policy hash active at the time of approval, and an on-chain transaction hash. Validation records are immutable and written to the ERC-8004 registry.

Constrained Delegation

EIP-7702-based delegation is always bounded by Brain policies. Delegation is visible on-chain and can be revoked by the user at any time. Agents can never exceed the limits defined in their policy — even if the backend were compromised, the smart account enforces limits independently.

---

For Developers

The Brain Protocol is accessible via the Brain SDK and a set of deployed smart contracts on Base and Ethereum.

import { BrainClient, PolicyBuilder } from '@brain-protocol/sdk';

const client = new BrainClient({ rpcUrl: '...', bundlerUrl: '...', signerKey: '...' });

const policy = new PolicyBuilder()
  .setSpendLimit('500 USDC/day')
  .setAllowedAssets(['USDC', 'ETH'])
  .build();

const agent = await client.agents.deploy({
  accountAddress: account.address,
  name: 'my-agent',
  metadataUri: 'ipfs://Qm...',
  policy,
});

Start building:

• Quickstart — deploy your first agent in five minutes

• Core Concepts — understand each layer of the protocol

• Smart Contracts — full contract interfaces and reference

• SDKs and APIs — TypeScript SDK documentation