AgentPay Core (USDC on Base) — OpenClaw Skill

What it is

AgentPay Core is a pay-per-call rail for OpenClaw skills using USDC on Base.

It provides a standard flow:

1. •Quote a price for an action
2. •Authorize (scoped, time-limited) usage
3. •Record usage + attach run/artifact proofs
4. •Produce verifiable receipts
5. •Enforce idempotency to prevent double-charge

Why it exists (agent pain)

• •Agents retry → double-charge risk without idempotency
• •Operators need spend limits + audit trails
• •Skill authors need metering + receipts to charge safely

Chain / Token

• •Network: Base (default)
• •Token: USDC
• •Dev mode: Base Sepolia + testnet USDC (until mainnet launch)

Minimal API (v0)

• •GET /v1/quote?sku=<action>&units=<n>
  • •returns {price_usdc, expiresAt, quoteId}
• •POST /v1/authorize
  • •body {quoteId, principal, scope, ttlSeconds, idempotencyKey}
  • •returns {authorizationId, token} (token is signed; scoped)
• •POST /v1/usage/record
  • •body {authorizationId, runId, unitsUsed, callRef, proofs:[...], idempotencyKey}
  • •returns {receiptId}
• •GET /v1/receipt/:receiptId
  • •returns receipt with signatures + pointers to run artifacts

Payment model (v0)

• •Prepay deposit USDC to a wallet address → system issues scoped authorizations until balance is exhausted.
• •(Later) on-chain per-call settlement or escrow.

Safety defaults

• •No transfers without explicit operator approval (until policies/limits are proven).
• •All calls require idempotency keys.
• •Receipts always include runId + evidence pointers.

TODO (next)

• •Reference TypeScript client + examples
• •Base Sepolia contract or ledger-backed escrow (v1)
• •Policies: per-principal spend caps; allowlists