Ant Digital Technologies introduced a four-layer infrastructure framework for autonomous agent transactions at Hong Kong Web3 Festival on April 20. CTO Yan Ying called the proposal the “4R Full-Stack Architecture,” covering runtime execution, payment rails, agent identity registry, and root infrastructure built on blockchain and zero-knowledge computation.
The Four Fault Lines
Yan Ying framed the architecture around four specific problems that current agent systems cannot solve by patching software alone, according to the announcement: execution failures from prompt logic vulnerabilities, an accountability vacuum caused by AI lacking verifiable identity, transactional barriers from payment gateways built for human principals, and collaboration risks when agents with no prior relationship need to establish mutual trust.
“This cannot be resolved by patching software,” Yan Ying stated. “It requires a ground-up redesign at the infrastructure layer.”
What Each Layer Does
The Agentic Runtime layer centers on DT Claw, Ant’s product embedding the CARLI safety model to enforce behavioral constraints at the execution level. It supports multi-model compatibility and financial-grade compliance, with every AI operation designed to be controllable, auditable, and recoverable.
Payment Rails provide native on-chain payment channels integrating agent-driven decision-making with verifiable credentials. For high-frequency micropayment scenarios, the platform uses Ant’s Jovay Layer2 network for sub-120-millisecond transaction confirmation and cross-chain settlement. The design targets a closed loop balancing financial-grade security with low developer and user adoption costs.
The Agent Registry assigns on-chain identities to each agent using the DID (Decentralized Identifier) standard and ERC-8004, making every instance of inter-agent collaboration traceable and verifiable.
Root Infrastructure combines blockchain with ZKVM (zero-knowledge virtual machine) technology, enabling off-chain computation verified on-chain. As Yan Ying put it: “Even two agents with no prior relationship can establish trust through code and transact with confidence.”
Ant’s Position in the Stack
Ant Digital Technologies spent a decade building financial-grade security, privacy computing, blockchain, and compliance infrastructure for Ant Group’s payment ecosystem. The 4R Architecture represents an attempt to repurpose that engineering depth for the agent economy, where AI systems hold assets and exercise transactional authority rather than simply generating text.
The timing coincides with a broader wave of agent payment infrastructure launches. MetaComp announced its Know Your Agent framework at Money20/20 Asia earlier this week, and Newborn Town launched NUSD Pay using the x402 HTTP protocol for machine-to-machine transactions on April 21.
The Infrastructure Convergence
Three companies proposing agent payment and identity infrastructure in the same week signals that the industry has moved past debating whether agents need financial infrastructure to competing over who builds it. The open question is whether blockchain-based approaches like Ant’s 4R stack or protocol-level solutions like x402 will win adoption among enterprise builders who need agents to transact autonomously.