Architecture

How VANTA routes AI agent intents through encryption, proof generation, and private execution.

Encrypted Intent Flow

When an AI agent wants to execute a strategy (e.g., swap, provide liquidity, rebalance), it never broadcasts a raw transaction. Instead, the flow works as follows:

The agent's strategy details (which token, how much, what price, timing) remain encrypted from submission to execution. Only the solver — selected by stake-weighted reputation — can decrypt the intent.

ZK Proof Circuit

VANTA uses Groth16 proofs on BN254 to verify that an encrypted intent satisfies protocol constraints without revealing the intent itself:

• Solvency proof — the agent's wallet has sufficient balance to cover the intent at submission time
• Validity proof — the intent conforms to protocol rules (valid action types, reasonable slippage, non-negative amounts)
• Commitment binding — the encrypted payload matches the proof's public input commitment (Poseidon hash)

MEV Shield

Even with encrypted intents, the execution layer can still leak value. VANTA's MEV Shield addresses this at two levels:

Network Components

$VANTA Token

The native token powering governance, privacy tier access, and protocol fee settlement.

Token Utility

• Governance — $VANTA holders vote on protocol upgrades, solver parameters, and treasury allocation via on-chain proposals
• Privacy Tier Staking — stake $VANTA to unlock higher privacy tiers with faster execution, dedicated solvers, and priority bundling
• Fee Payment — protocol fees for intent submission and execution are paid in $VANTA (with SOL auto-conversion available)
• Solver Bonds — solvers must bond $VANTA as collateral; slashed if they fail to execute or attempt manipulation

Tokenomics

Staking Tiers

SDK Reference

TypeScript SDK for building private AI agents on VANTA Protocol.

Installation

npm install @vanta/sdk

Or with yarn / pnpm:

yarn add @vanta/sdk
pnpm add @vanta/sdk

Quick Start

Initialize an agent and submit an encrypted intent:

import { VantaAgent } from '@vanta/sdk';

const agent = new VantaAgent({
  rpc: 'https://api.mainnet-beta.solana.com',
  privacyLevel: 'maximum',
  mevShield: true
});

// Submit an encrypted swap intent
const result = await agent.submitIntent({
  action: 'swap',
  input: { mint: 'SOL', amount: 10 },
  output: { mint: 'USDC' },
  encrypted: true
});

console.log('Intent ID:', result.intentId);
console.log('Privacy score:', result.privacyScore);

Configuration

interface VantaConfig {
  /** Solana RPC endpoint */
  rpc: string;
  /** Privacy level: 'standard' | 'enhanced' | 'maximum' */
  privacyLevel: 'standard' | 'enhanced' | 'maximum';
  /** Enable Jito MEV Shield routing */
  mevShield: boolean;
  /** Custom relay endpoint (optional) */
  relayUrl?: string;
  /** Wallet keypair or adapter */
  wallet?: Keypair | WalletAdapter;
  /** Network: 'mainnet-beta' | 'devnet' */
  network?: string;
}

API Reference

Contracts & Security

On-chain programs, audit status, and security disclosures.

Solana Program

Audit Status

Security Scope

• On-chain program — intent settlement, proof verification, bond management, treasury operations
• ZK circuits — Groth16 circuit soundness, Poseidon hash integrity, proof generation correctness
• Relay infrastructure — encrypted intent handling, solver selection, bundle submission
• SDK — client-side encryption, key management, proof generation

Security Model

Bug Bounty

VANTA runs an active bug bounty program for responsible disclosure of security vulnerabilities.

Report vulnerabilities to security@usevanta.xyz with a detailed description and reproduction steps.

FAQ

Frequently asked questions about VANTA Protocol.