Turnkey has launched Agentic Payments, a production-ready infrastructure layer enabling AI agents to transact onchain with non-custodial wallets and policy controls.
US-based Turnkey has announced the launch of Agentic Payments, a production infrastructure product that provides AI agents with the core payment stack required to transact onchain autonomously. The solution combines non-custodial wallets, transaction lifecycle management, gas sponsorship, real-time balance visibility, and policy controls within a single integrated system, removing the need for development teams to assemble these components separately.
The product is designed for developers building AI agents that need to move value onchain, including teams working on agentic treasury systems, automated disbursements, agent-to-agent settlement, subscription billing, marketplace payouts, and other autonomous financial workflows.
Core capabilities and security model
Agentic Payments is built around four capabilities. Each agent receives a dedicated non-custodial wallet without accessing raw private keys. Turnkey generates and stores keys inside secure enclaves, a type of Trusted Execution Environment, with agents authenticating via API key to request signatures. Transaction lifecycle management covers the full flow from construction through broadcast and status tracking in a single API call, with gas sponsorship removing the need for agents to hold native tokens for transaction fees.
Real-time balance visibility allows agents to check available funds before initiating payment decisions, reducing failed or incomplete transactions. Policy controls evaluate every signing request inside the enclave before a signature is produced, scoping agent activity by recipient address, contract, function, chain, and value limit. High-value actions can require co-approval from a human or another agent.
The enclave-level policy enforcement means that even a compromised application backend cannot produce a transaction that violates defined policy - a distinction from application-layer controls that can be bypassed if the application itself is compromised.
Use cases and production deployment
Turnkey cites several production use cases: AI-powered payroll and disbursement engines, agent-to-agent settlement in multi-agent commerce, automated subscription billing with wallet-level spend limits, marketplace settlement between participants, and automated treasury operations including fund rebalancing and sweeping within human-defined policies.
Catena, which is building agentic treasury and banking infrastructure, is cited as a production user of the system. Its implementation requires approval across the customer, the customer's agent, and the Catena treasury agent for transactions, with policy enforcement handled at the enclave level rather than the application layer.
The launch reflects a broader shift in agentic AI from recommendation-making to economic action - a transition that requires payment infrastructure capable of supporting autonomous value movement within defined constraints, rather than requiring human approval for every transaction.
