MoonPay launched the MoonAgents Card on May 1. It is a virtual Mastercard built so AI agents can spend stablecoins at any merchant on the network. The card is issued by Monavate Ltd through Baanx Group, which is the same Mastercard issuing stack behind several other crypto debit products. The Block and crypto.news both confirmed the UK and LATAM launch, with US and EU listed as "coming months."
The card itself is not the news. Stablecoin debit cards have existed for years. The funding model is what makes this product new: there is no pre-loaded balance and no custodial top-up. The user delegates spending authority to a smart contract. At the moment a transaction authorizes, Monavate pulls the exact stablecoin amount from the user's onchain wallet. If the merchant declines, the funds never leave the wallet. Approvals can be revoked in a single CLI command.
Why a Card Built for Software Behaves Differently
Every other crypto card assumes a human is the one tapping. Even the cards aimed at developers expect a human to fund an account, the account to hold the balance, and the card to draw from the account. The MoonAgents Card breaks that model. The card is meant to be issued, funded, and frozen by code.
MoonPay's CLI has processed over 4 million tool calls since launch, per the announcement. The first million took thirty days. The second million took seven. That curve is what the product is built around. If millions of software agents are already querying prices and signing transactions, the gap they cannot close today is paying a flight, an API bill, or a household subscription. The card closes that gap.
The customer profile changes too. The press release quotes Ivan Soto-Wright on agents transacting "at machine speed across millions of merchants" and JP Richardson of Exodus saying that the next wave of wallets and cards "looks nothing like what exists today." That is not consumer cashback positioning. That is infrastructure aimed at developers, finance ops teams, and the agent economy that MoonPay, Exodus, and the Open Wallet Standard consortium are betting will be real in 18 months.
How the Card Actually Works
The flow is CLI-first. Five commands after npm install -g @moonpay/cli:
| Step | Command | What it does |
|---|---|---|
| Login | mp login then mp verify <code> | Browser sign-in to MoonPay account |
| KYC | mp card onboarding start with name, country, DOB, phone | Kicks off Baanx ID verification |
| Address | mp card onboarding finish with address and --accept-terms | Submits residency, accepts T&Cs |
| Issue | mp card create | Returns PAN, CVV, and expiry instantly |
| Delegate | mp card wallet link --wallet <name> --currency usdc --amount 5000 | Approves smart contract to pull up to 5,000 USDC |
The two commands that matter for security are mp card freeze, which pauses every transaction without touching the wallet, and mp card wallet unlink, which revokes the smart contract approval entirely. Both are documented on the official launch page.
A few details that the wires did not flag are worth pulling out. The launch references USDC and Solana wallets in the example flow. Other chains and stablecoins are not yet listed, which is consistent with MoonPay's existing rails but narrows the product for now. There is no cashback, no yield, and no staking. That is unusual for a 2026 crypto card. Crypto.com sits between 1 and 5 percent, Kraken goes up to 5 percent, and the Jupiter Global card ladders from 4 to 10 percent. MoonAgents has zero rewards, which is a positioning decision rather than an oversight.
The product page advertises "zero-fee stablecoin onramps." Network spread, FX, and chain gas are not addressed in the launch material. As we covered in the USDC spread vs gas analysis, the disclosed fee on a stablecoin card is rarely the full cost. Identity verification runs through Baanx with passport and selfie, and country of residence is required at onboarding, so a US passport holder cannot pick up the card today by claiming a UK address. Baanx checks both.
SpendNode's Take: This Is a B2B Product Wearing a Consumer Wrapper
The headlines are about AI agents shopping for coffee pods. The actual play is corporate spending automation, dev-ops budgeting, and treasury operations for crypto-native companies that want a Mastercard rail wired into a multisig.
A few things point that direction. The product ships as a CLI rather than an app, which is not how consumer cards launch. The CLI usage curve (4 million calls, doubling in a quarter of the time) is a developer adoption pattern. And the absence of any rewards program in a market saturated with Mastercard-network crypto cards tells you the customer is not a retail user comparing this product to the Crypto.com tier ladder.
If that thesis is right, MoonPay benefits because they sell more CLI usage and onramp volume. Monavate and Baanx benefit because they get another distribution surface. Mid-size crypto companies benefit too, because they get programmable card-issuing without going through the long process of becoming a Mastercard principal member directly. The losers are the agent-card products that built on a custodial pre-load model. They now have to argue why a user should hand over funds in advance when MoonPay can pull at the swipe instead.
What Card Users and Developers Should Actually Do With This
For a typical reader of this site, the MoonAgents Card is not a substitute for a self-custodial card like Gnosis Pay or Tria. Those products give you onchain settlement plus rewards, and they are issued in the EU and the UK without a CLI dependency. If you want to spend stablecoins from your own wallet and earn something back, those are the products to compare.
For a developer or finance ops person at a crypto company, the calculation is different. The MoonAgents Card lets you put a card in front of an agent that pays AWS bills, renews data subscriptions, or executes a recurring buy at a vendor that does not accept stablecoins natively. The cap is set per-wallet. The ledger is auditable through mp card transaction list. Revocation is instant. That is operationally tighter than handing a corporate card to an automation script.
A practical first move: install the CLI on a test machine, KYC if you have a UK or LATAM address, link a small amount, and run a real merchant transaction before connecting any LLM to it. We would start with 100 USDC, not the 5,000 in the example. Prompt injection on agent cards is a live attack surface, and the cap is the worst-case loss number until that surface is exercised in the wild.
For everyone else, bookmark the product page and wait for the US and EU rollout. The question for SpendNode readers in those regions is whether MoonPay will keep the card cashback-free at launch, or whether competitive pressure forces a rewards add-on once the audience expands beyond developers.
Overview
MoonPay's MoonAgents Card is the clearest signal yet that crypto card infrastructure is splitting. One product line is consumer cashback cards competing on rewards rates. The other is programmable card-issuing aimed at agents, developers, and crypto-native treasuries. MoonAgents has no cashback, no app, and no consumer surface, and that is the point. Whether the smart-contract-pull funding model becomes the new default for agent cards over the next two quarters depends on how fast the custodial pre-load incumbents respond.
