Recall

How Your Agent Recalls Its Payments

An AgentPay agent can ask the rail "what have I paid, to whom, and for what?" — and get an instant, private answer drawn from its own ledger. No blockchain queries, no other agent's data, ever.

The recall loop

Ask — the agent calls its recall endpoint with its own token. The token identifies the agent and authorizes only its own records.

Scope — the rail returns only rows where this agent is the payer or the payee. Strict isolation is enforced at the query — there is no way to see another agent's payments.

Recall — the agent gets back its settlements with amounts, counterparties, what was bought, and the tx_hash for each, ready to reason over.

What the agent gets back

{
  "settlements": [
    {
      "amount_usdc": 0.10,
      "payee": "cipher-agent",
      "resource": "on-chain DeFi data",
      "tx_hash": "0x9f3a…",
      "created_at": "2026-06-30T14:02:11Z"
    },
    ...
  ]
}

Each line is both a memory and a receipt: the agent remembers what it paid for, and the tx_hash lets anyone verify it on Base.

From recall to relationships

Once an agent can recall its payments, the next step is meaning. AgentPay derives a payment graph from the ledger — grouping by counterparty, frequency, and category — so an agent can answer richer questions:

"Which agents do I rely on most?"
"How much have I spent on market data this month?"
"Who are my most trusted counterparties?"

This turns "agents that pay each other" into "agents that know who they pay, trust, and depend on" — the foundation for smarter autonomous commerce.

Private by construction. Recall is always scoped to the requesting agent. One agent's spending history is never visible to another — see Private Payments for the full isolation model.

How x402 escrow works →

🔒 Privacy 📒 Ledger 🧠 Recall 🤝 Escrow 📚 Docs