MoonPay, a fintech that specializes in crypto on and off-ramp service, has launched the Open Wallet Standard – an open-source protocol that provides AI agents with a secure, universal way to hold value, sign transactions, and pay for services across multiple blockchains.
According to the company, the standard aims to bridge a gap in how AI agents currently interact with crypto systems. Existing protocols that enable machine-driven payments focus on transaction rails, but fail to address how wallets and private keys are managed across tools and networks.
Today, agents often operate across fragmented systems, where each application maintains its own keys and balances separately, resulting in inefficient processes and heightened security vulnerabilities. A lack of coordination in the key management process can expose funds to hacks or losses, particularly if agents operate in environments with varying security protocols.
MoonPay’s Open Wallet Standard Creates a Shared Layer for Agentic Commerce Across Blockchains
MoonPay’s Open Wallet Standard replaces that fragmentation with a shared layer, where agents can operate from a single pool of funds across multiple protocols and chains rather than managing multiple disconnected accounts. The company aims to counter the supposed risks by providing a cohesive framework for wallet access and transaction execution, benefiting both AI developers and their end users.
The framework currently supports Bitcoin, Ethereum, Solana, all Ethereum-compatible networks, and other blockchain ecosystems.
As these agents become more prevalent across trading, e-commerce, and automated financial services, the demand for seamless interactions with blockchain ecosystems and technologies will continue to grow, and MoonPay aims to meet that.
Key management is the central focus of MoonPay’s agentic wallet architecture. The standard stores private keys in an encrypted local vault and briefly decrypts them to sign a transaction before clearing them from memory. Meanwhile, the signing process runs in an isolated environment, keeping the keys out of the agent’s active runtime. It also includes policy controls that let operators set spending limits and transaction restrictions before any approval is granted.
MoonPay CEO and co-founder Ivan Soto-Wright said the agent economy did have payment rails but lacked a wallet standard. He described the Open Wallet Standard as the completion of that infrastructure stack, and it is not intended to compete with existing protocols but to make them more useful by providing a common wallet for any protocol that requires a signed transaction to call.
Open Wallet Standard Builds on MoonPay’s x402-Compatible Non-Custodial Software Layer
The standard builds on MoonPay Agents, a non-custodial software layer released by the company in February 2026 that allows AI agents to interact with wallets and execute transactions autonomously via the MoonPay CLI. That product came with x402 compatibility – the internet-native payment protocol that lets web services, APIs, and apps request micropayments before serving content, with payments often handled in stablecoins directly over standard HTTP.
Earlier this month, MoonPay integrated support for signing transactions with a Ledger hardware wallet into the agent stack, making it the first agentic wallet to support hardware-backed transaction approval. But when the company faced the issue of having no shared standard for how agents interact with wallets, it took the infrastructure behind MoonPay Agents and generalized it across different blockchains and runtimes, culminating in the Open Wallet Standard released under an MIT open-source license for public use.
More than a dozen companies, spanning across the crypto and fintech industries, have contributed to the standard, including PayPal, OKX, Ripple, Tron, TON Foundation, Solana Foundation, Ethereum Foundation, Base, Filecoin Foundation, Dflow, Uniblock, Dynamic, Allium, Simmer.Markets, and Circle.
BitGo, Coinbase, and Tempo Explore Machine-to-Machine Commerce
Meanwhile, on Monday, BitGo launched a Model Context Protocol server connecting its custody and infrastructure platform to AI-native development tools through natural language. This integration allows large-language models (LLMs) such as ChatGPT and code editors to retrieve wallet documentation, API references, and staking system information directly within developer workflows.
Other companies actively exploring agentic commerce include Coinbase, whose x402 protocol enables stablecoin transfers over HTTP for AI agents and applications, and Visa and Stripe-backed Tempo network, which released tools last week allowing AI systems to initiate and execute payments without human interference.
