Introduction

ZenithPay is a spend governance layer for AI agents on Base.

It combines:

• Privy-managed agent wallets (secure key custody)
• SpendPolicy.sol on Base (hard policy gate)
• x402 payment protocol on Base
• ledger + approvals for full auditability

Why ZenithPay exists

AI agents can spend money faster than humans can supervise them. A secure payment stack needs:

• hard limits that cannot be bypassed by backend bugs
• deterministic payment routing for machine-to-machine APIs
• clear operator visibility for approved/blocked/pending outcomes

ZenithPay gives all three with onchain policy enforcement first, then payment execution.

System components

1. Agent Wallet
   • Privy-managed wallet, no raw private key handling in app code.
2. Policy Engine
   • SpendPolicy.sol enforces per-tx and daily constraints on Base.
3. Payment Engine
   • POST /pay orchestrates check -> x402 settlement.
4. Seller Route
   • /sell/agent-intel implements x402 seller flow (402 challenge, verify, settle, paid response).
5. Control Plane
   • approvals queue, ledger, and MCP tools for operational control.

Payment lifecycle (high level)

1. Agent submits POST /pay with serviceUrl, maxAmount, and intent.
2. ZenithPay checks SpendPolicy first.
3. ZenithPay executes x402 verify/settle on Base.
4. ZenithPay writes ledger and returns approved, pending, or blocked.

Current production references

• API: https://zp-api-q4hd.onrender.com
• Seller route: https://zp-api-q4hd.onrender.com/sell/agent-intel
• Chain: Base mainnet (8453)
• USDC: 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913

Minimal pay example

curl -X POST https://zp-api-q4hd.onrender.com/pay \
  -H "Authorization: Bearer $ZENITHPAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agentAddress": "0xcadf...1a9",
    "serviceUrl": "https://zp-api-q4hd.onrender.com/sell/agent-intel",
    "maxAmount": "0.01",
    "intent": "Fetch paid seller intel"
  }'