Who's behind the agent? Security, trust, and compliance in agentic payments
We’ve been collaborating with others to explore how agentic commerce and enterprise agents will work. Alfonso Gómez-Jordana Mañas, co-founder of Crossmint shares his thoughts on the current state of payment security and how compliance mechanisms like KYC and AML need to evolve to support agentic commerce.
OpenClaw provided users with an open-source framework to launch AI agents. Practically overnight, thousands of builders created millions of agents performing tasks from managing inboxes and scheduling meetings to making purchases and executing trades.
However, agentic payments infrastructure wasn’t developed. Without a secure, universal payment standard for AI agents, developers resorted to ad-hoc workarounds: they embedded raw credit card numbers and CVCs directly in agent instructions, or handed agents private keys to crypto wallets. It worked, but was risky and there wasn’t a compliance layer with identity verification, sanctions screening, or AML controls in place.
This is why we built lobster.cash, an open payment solution for AI agents that ensures security, compliance, and universal support for payment methods. While lobster.cash runs on Crossmint’s existing stablecoin and wallet infrastructure, we also made sure to work with leading partners like Visa, Mastercard, Solana, Circle, Basis Theory, and Stytch. The solution will be kept open source to ensure that it is a trusted standard that is accessible to all.
lobster.cash creates flexible payment security for agents
Most initial attempts at agentic commerce hit a block when merchants and card networks flagged agent-initiated transactions as bot activity. Even when builders found workarounds, there wasn’t a way to limit access to card details or enforce controls like merchant restrictions or approval thresholds.
The lobster.cash architecture addresses this directly. Agents get access to virtual credit cards and USDC stablecoin wallets through a single integration, but they never receive raw card data or the private keys. The human account holder defines the guardrails, including per-transaction limits, daily spending caps, merchant allowlists, and approval requirements above certain thresholds. The agent can only spend when it has explicit authorization to do so.
Lobster.cash solves the bot-flagging problem by enabling agentic cards via Visa Intelligent Commerce and Mastercard Agent Pay. Both programs are built specifically to help card networks and merchants recognize and authorize legitimate agent-initiated transactions.
Offering the flexibility to support crypto payments is also important. Stablecoin transactions enable global payments that settle quickly, cost fractions of a cent, and work around the clock, which is ideal for high-frequency payments that agents often make (API calls, compute, data access).
Built-in compliance helps developers launch
Adding controlled payment authorization to agents solves one problem and introduces a new one.
If you’re issuing a new card and crypto wallet for an agent, you likely need to comply with Know Your Customer (KYC) and anti-money laundering (AML) regulations. This could mean collecting and verifying the user’s identifying information; screening against sanctions lists, PEP databases, and adverse media; and performing ongoing monitoring.
For many teams building agentic products, these regulatory challenges can stall projects. AML compliance touches onboarding, transaction monitoring, dispute resolution, and regulatory reporting. Building a complete solution means integrating ID verification, sanctions databases, case management tooling, and reporting workflows, and maintaining all of it as regulations shift across jurisdictions. It's months of engineering time spent on a problem that isn't your core product.
This is why we chose Persona as our KYC and AML provider. Here’s how it works:
Agent transactions that require identity verification or compliance checks are automatically routed through Persona, which handles sanctions screening, watchlist monitoring, and risk-based decisioning across our platform.
Crossmint customers get compliance embedded in their infrastructure layer instead of building and operating it themselves.
The practical effect is that developers who integrate Crossmint don’t need to stand up a sanctions screening pipeline. Instead, they can connect their agents to Crossmint’s API to give the agents payment capabilities, verify users, screen transactions, and manage identity data to help meet global AML requirements.
The road ahead
Agentic commerce is past the demo stage, but the infrastructure is still maturing. Today, agents can operate under the identity of a human account holder, which is a workable model when the agent is a delegate constrained by human-defined permissions.
Over time, the industry will need to develop credentialing and attestation standards for agents, which may include verifying what model it runs, who deployed it, and what permissions it holds. As that happens, identity verification will extend to the agent with what’s quickly been dubbed Know Your Agent (KYA) protocols.
In addition, as agent-initiated transactions grow, regulators may expect the same compliance rigor applied to agentic commerce as to any other payment channel. Teams that build on infrastructure designed with compliance in mind from day one won't have to retrofit it later.
At Crossmint, we've built the payment layer for agentic commerce, including wallets, stablecoin rails, virtual cards, programmable spending controls, and more. By partnering with providers like Persona for human identity and compliance, we've made it possible for developers to launch agentic payment products without having to become compliance experts in the process.
If you're building agents that need to transact, reach out to Crossmint to explore our agentic payments solutions. And if verifying the humans behind agents and meeting compliance are on your roadmap, get in touch with the team at Persona.