OKX Launches Payment Protocol for Autonomous AI Agents
AI agents are evolving from responding to instructions to carrying out tasks and business processes. Most existing payment tools, however, handle only one-off payments and lack functions for quoting, negotiation, escrow, settlement and dispute resolution. Crypto exchange OKX is using on-chain infrastructure to enable software agents to purchase services and transact with one another autonomously, with less human intervention.
OKX released its open-source Agent Payments Protocol (APP) on April 29, 2026, for implementation on blockchains including Ethereum and Solana. Its Payment SDK already supports one-off, batch and usage-based payments on X Layer, with zero gas fees, while escrow functionality has yet to launch. The accompanying self-custodial Agentic Wallet supports more than 20 blockchains and is protected by a trusted execution environment, or TEE. In an official example, 10 USDC is prefunded so an agent can pay for hundreds of API calls.
All Coverage
1 original reportsThe Backstory
The history behind this eventOKX Enters Crypto AI Agent Race With OnchainOS Upgrade
OKX is positioning OnchainOS as a unified execution framework for autonomous crypto trading agents, combining wallet infrastructure, liquidity routing and onchain data. The move reflects exchanges' efforts to advance AI beyond analytics and into automated systems that can execute onchain operations directly. This lowers barriers for developers but increases the importance of key management and transaction risk controls.
OKX formally entered the crypto AI agent platform race with an AI upgrade to OnchainOS in July 2026. The new version supports more than 60 blockchains, allowing developers to connect wallets, cross-chain liquidity and real-time onchain data through a single toolkit. This is designed to accelerate the development of bots capable of making autonomous decisions and executing trades.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.