Agent finance map

Agent Capital

The reference on how autonomous AI agents hold, move, and manage money: wallets, budgets, x402 payments, treasury, controls, and the risk of agents transacting with agents.

Verdict map

1

Wallet: Who or what can sign?

2

Policy: What is the agent allowed to do?

3

Rail: How is value moved?

4

Treasury: How is spend reconciled?

5

Risk: What breaks under autonomy?

Key facts

What this page establishes

  • Agent capital is the stack around agent money: wallet, policy, rail, treasury, and risk controls.
  • The safest agent wallet is usually a delegated payment tool with small balances and explicit limits.
  • Spend controls should be enforced before signing or authorization, not only observed after settlement.
  • Free tools on this site help estimate budgets, draft spend policies, and model x402 cost envelopes.

Short answer

Can AI agents have their own money?

They can be given access to money-moving tools, wallets, payment credentials, and budgets. That is different from legal ownership. In practice, an "agent wallet" usually means a wallet or payment credential controlled through software, with the agent allowed to request or trigger actions inside limits set by a human, company, protocol, or custody provider.

The safest framing is operational: an agent should hold the smallest useful balance, receive the narrowest useful permission, and produce the richest useful audit trail. Wallets and payment rails matter, but the core product is spend control.

Citation-ready claims

Decision frame

The market is splitting into three layers.

Wallet infrastructure

AgentKit, Agentic Wallet, server wallets, MPC custody, smart accounts, and payment credentials decide what can sign.

Payment protocols and networks

x402, AP2, ACP, card issuing, and network products decide how an agent proves intent and completes payment.

Finance operations

Spend policy, audit logs, reconciliation, approvals, disputes, and refunds decide whether agent spend can be trusted at scale.