MoonPay has published an open-source wallet specification designed to let autonomous AI agents hold funds and execute transactions across multiple blockchains, addressing a key gap in how software agents interact with crypto infrastructure.
The standard defines a common interface so agents and developer tools can access a single wallet across chains and platforms instead of each tool managing separate keys and balances. That unified model lets agents draw from one pool of funds rather than juggling multiple disconnected accounts.
Researchers at MIT Sloan have noted that AI agents increasingly rely on standard building blocks—APIs, messaging, payment actions and web access—to interact with people and other agents. MoonPay said earlier efforts mainly targeted payment rails while leaving wallet and key management unresolved, creating fragmentation and security concerns.
MoonPay’s design keeps private keys inside an encrypted local vault and performs transaction signing in a separate, isolated process so keys do not run inside an agent’s runtime. The framework also includes policy controls that let users set spending limits and other constraints; transactions must satisfy those policies before being approved and signed.
The specification is modular and open source, with separate components for storage, signing, policy enforcement and chain adapters. MoonPay has published the code and packages on common developer platforms including GitHub, npm and PyPI, and said more than a dozen companies contributed to the specification. Named participants include PayPal, OKX and Circle alongside several blockchain foundations and infrastructure providers.
Founded in 2019, MoonPay offers infrastructure to move money between fiat and digital assets, providing on- and off-ramps, trading and crypto payment services around the world.
The release comes amid broader industry activity to make AI agents first-class economic actors. BitGo, a digital-asset custody and infrastructure firm, recently announced a Model Context Protocol (MCP) server to give AI-driven tools natural-language access to wallet functions, transaction flows and staking features. That integration aims to surface documentation, API references and product information in-context inside AI-native workflows and developer environments.
Other initiatives include Coinbase’s x402 protocol for stablecoin transfers over HTTP, which targets APIs, apps and agent integrations, and projects from Visa and Tempo (backed by Stripe) that enable programmatic payment initiation by AI systems. Together, these developments reflect a push to let software interact directly with financial rails without relying solely on traditional user interfaces.
As with any reporting, readers should independently verify technical and security details before adopting new infrastructure.