Skip to main content
An Agent Wallet is a non-custodial smart wallet that an agent can spend from within limits the owner sets. For consumer agents, the user owns the wallet and adds the agent as a signer with a spend cap, allowed counterparties, and a time window. For autonomous agents with no end user, your backend owns the wallet directly. Permissions are enforced onchain and the owner can revoke them at any time. Neither you nor Crossmint takes custody of funds. Agent Wallets are how agents pay other machines: APIs behind x402 and MPP, other agents, and any contract onchain. For the problem this solves, read How Agents Pay.

​
Key Features

Non-custodial

The user, or your backend, holds the root key; nobody else can move funds

Scoped agent access

Spend cap, allowed counterparties, and time window per agent

Enforced onchain

A transaction outside the delegation is rejected at the wallet level

Machine-to-machine ready

Pay x402 and MPP endpoints per call, in stablecoins

Fund with a card

Users top up with the Crossmint onramp or a direct USDC transfer

Revocable

Remove the agent signer at any time; the user keeps the wallet and funds

​
When Are Agent Wallets the Best Fit?

  • The agent pays machines, not stores. APIs behind x402 or MPP and other agents settle in stablecoins, per call, with no merchant account.
  • Payments are small or frequent. Card fees and chargebacks make micropayments impractical; stablecoins do not have that problem.
  • The agent acts onchain. Sending stablecoins, swapping, bridging, and calling contracts all run from the same wallet.

​
How It Works

1

The user creates and funds a wallet

The Crossmint SDK creates a non-custodial wallet with email or passkey recovery. The user funds it through the Crossmint onramp or a direct USDC transfer.
2

The user authorizes the agent

The agent’s key is added as a signer, bound to a permission set the user approves: spend cap, allowed counterparties, and time window.
3

The agent transacts within scope

The agent signs and submits transactions with its own key, with no user prompt per action. Anything outside the delegated scope is rejected at the wallet level.
4

The user revokes access when done

Removing the agent signer ends its access immediately. The user keeps the wallet and its funds.

​
Two Wallet Models

​
Get Started

Quickstart

Run the reference app: create a user wallet, authorize an agent, and send a stablecoin transfer from your backend

Try the live demo

See the full flow without setting anything up locally

​
Guides

Create a User Wallet

A non-custodial wallet owned by your end user

Authorize the Agent

Grant your agent delegated, scoped control

Onramp & Add Funds

Fund the wallet with a card or a direct transfer

Server Agent Wallets

Backend-only wallets for autonomous agents

Using the Wallet

Send, swap, bridge, and call any contract

x402

Pay x402-protected endpoints per call, in stablecoins

MPP

Pay Machine Payment Protocol endpoints from an EVM wallet

Remove Agent Access

Revoke the agent signer from a user’s wallet

​
FAQs

No. The wallet is non-custodial. The user (or, for server agent wallets, your backend) holds the root signer, and the agent only holds a delegated key bound to the permissions the owner approved.
The spend cap, the addresses or protocols the agent can interact with, and the time window in which the delegation is valid. The limits are enforced onchain, so a transaction outside them is rejected at the wallet level.
Yes, at any time. See Remove Agent Access.
Agent Wallets are Crossmint smart wallets. The Wallets documentation covers chains, signers, recovery, and the full SDK.