MoonPay has introduced PayBox, a “payment vault” designed to let AI assistants such as ChatGPT and Claude execute crypto actions inside a conversation—while keeping users in control of their wallet permissions. The company positions the product as a safer way for AI to perform tasks like swapping tokens, bridging assets, and interacting with DeFi, using natural-language instructions from the user.
In practical terms, users connect a crypto wallet and payment methods to the AI assistant. The assistant then prepares transactions—such as token swaps, cross-chain transfers, or DeFi calls—based on what the user asks. MoonPay says approval can be handled via a passkey or through spending limits that allow the AI to carry out certain actions automatically within predefined boundaries.
Key takeaways
- PayBox aims to put consent first, offering per-transaction approval or limited autonomous execution based on user-set permissions.
- MoonPay says it protects wallet keys using multi-party computation and trusted execution environments to reduce direct access by both the AI assistant and MoonPay.
- PayBox supports multiple payment rails, including debit cards, bank accounts, Apple Pay, and PayPal, in addition to crypto wallets.
- The vault integrates with AI apps via an SDK, allowing developers to embed PayBox functionality into their own assistant experiences.
- x402 momentum continues, with public dashboards showing large recent transaction volumes across participating services.
PayBox: AI-driven crypto payments with user-controlled permissions
MoonPay’s central pitch with PayBox is control. The platform is built around the idea that an AI assistant should be able to request or construct crypto transactions from within a chat, but that the user retains authority over how those actions happen.
MoonPay says users can require approval for every transaction, or choose an approach where the AI is allowed to act automatically only within limits the user defines. Those limits can function as guardrails for autonomy—constraining what the assistant can do without additional confirmation.
To address key management and custody concerns, MoonPay states PayBox uses multi-party computation and trusted execution environments. The goal is to prevent either the AI assistant or MoonPay from independently accessing user funds. While the user initiates actions through the AI interface, the underlying design is intended to reduce the risk of unilateral fund movement.
From swaps to bridges: where PayBox fits in an AI workflow
PayBox is positioned for common “agent” behaviors in crypto—actions that are often difficult for users to execute safely or quickly. MoonPay says the system can support transaction types including:
- Token swaps initiated from chat prompts
- Cross-chain bridging and transfers across networks
- DeFi interactions constructed as transactions based on user intent
MoonPay also highlights that PayBox works with multiple blockchains and multiple payment methods. According to the company, users can combine crypto wallets with fiat-to-crypto and payment rails such as debit cards, bank accounts, Apple Pay, and PayPal.
For developers, MoonPay indicates the product can be integrated through a software development kit (SDK), suggesting a strategy beyond serving end-users directly and instead enabling other AI-driven apps to embed transaction authorization and execution flows.
PayBox’s parallel track: x402 and the push for AI-native payments
PayBox also supports x402, an open payment protocol originally developed by Coinbase. x402 is intended to enable AI agents to make internet-native payments, aligning with the broader trend of “agentic” applications that can transact without requiring users to manually navigate payment steps.
In April 2026, the x402 protocol was contributed to the Linux Foundation. The Linux Foundation said the protocol is now governed as an open, vendor-neutral industry standard, through the launch of the x402 Foundation. Earlier coverage from Cointelegraph also noted growing ecosystem activity around x402.
Why the x402 ecosystem growth matters
Coinbase has continued expanding the x402 toolset. In June, the crypto exchange launched features aimed at helping AI agents accept USDC payments, trade crypto, discover paid services through an AI marketplace, and process high-frequency micropayments more efficiently—capabilities designed to reduce friction for automated transactions.
Other industry participants have moved in the same direction. Amazon Web Services integrated x402 into its Bedrock AgentCore Payments service, while Fireblocks launched an x402-compatible payments framework for AI agents and joined the x402 Foundation. Together, these efforts point to a broader attempt to standardize how AI agents receive payment instructions and complete transactions.
Network activity is another indicator that adoption may be accelerating. According to Chainalysis, agentic payments on Coinbase’s Base network surpassed 100 million transactions within roughly nine months, as reported in a June 3 report. Chainalysis also noted early usage could have been influenced by speculative applications, underscoring that transaction volume alone doesn’t always map directly to stable commercial demand.
Still, the public x402scan dashboard shows more than 12.7 million transactions over the past 30 days across participating services, indicating that activity continues to build across the ecosystem.
What to watch next
As PayBox and x402-related infrastructure mature, the key question for users and builders is whether safety controls and key protection can scale smoothly alongside growing transaction throughput—so far, the trend suggests more AI payment workflows are moving from experiments into repeatable, permissioned execution. The next signal to monitor will be how widely these tools are adopted across real assistant experiences and whether transaction growth translates into robust, non-speculative usage.
