The architectural pieces have been described in the abstract across the preceding parts. This part walks the production-grade examples that put them together and the open questions that will shape the resource's tracking through the next twelve months.
Where it appears in production
The x402 protocol plus USDC is the most concrete current production example of an end-to-end agentic payment loop on a tokenised rail. The pattern is end-to-end: an API service returning HTTP 402, an agent client that interprets the request, signs and broadcasts the payment in USDC, and retries with proof. Developer-tooling and AI-infrastructure startups have integrated against this through 2024 to 2026; volume is small relative to traditional payment flows but integration cost is low and the pattern is generalisable to any chain that hosts USDC or another payment stablecoin.
Programmable conditional payments via Project Ensemble in hong kong and the MAS Project Orchid PBM work in Singapore are the regulator-coordinated examples in a tokenised deposit rather than stablecoin context. Ensemble's tokenised deposit layer supports conditional transfers; Orchid's PBM structures contemplate programmable conditions on transferability. Both are production-grade in pilot; broader rollout depends on the operating banks integrating the patterns into client-facing products.
Skyfire and Catena Labs are the architectural sketches to know but not yet bet on. Skyfire's virtual-card model is the simpler integration today, layered on existing card networks. Catena's chain-native wallet infrastructure is the more ambitious bet, with the corresponding execution risk.
A worked hypothetical that ties the threads together. A corporate treasurer authorises an agent to route incoming USDC into BUIDL whenever the treasury wallet exceeds a threshold, with the BUIDL position liquidated and converted to a tokenised deposit at a Singapore-licensed bank when an outgoing payable falls due. The corporate is the principal, KYC'd by both the BUIDL transfer agent and the bank. The agent operates under express authority documented at onboarding, with verifiable credentials demonstrating it on demand. The USDC inbound is chain-level transferable, subject to issuer blocklist screening. The BUIDL subscription runs through a whitelisted wallet under the principal's KYC. The tokenised deposit issuance runs through the bank's chain, with the agent's transfer operating under the bank's whitelist. Three rails (public chain for USDC, permissioned chain for BUIDL, bank chain for the deposit), each within its own regulatory perimeter, with the agent as the routing layer above. Most jurisdictions accommodate this. The flow that does not work in 2026 is the same loop with the deposit leg in wholesale CBDC (wCBDC), because the corporate treasurer cannot hold wCBDC.
This is an architectural sketch, not a volume claim. The operational and legal pieces exist; the flow has not been productised at scale because regulatory clarity, operational policy, and agent-stack maturity are still settling.
Open questions for the resource
Which regulator is first to address agentic flows specifically, with text rather than inference. The candidates are MAS, with its Project Orchid and DTSP work positioning it as the most agent-curious of the major APAC regulators; the Bank of Korea (BoK) and Financial Services Commission (FSC), with the Phase 2 Digital Asset Basic Act under south korea still in legislative progress and capable of incorporating agent provisions; and the European Banking Authority (EBA) under MiCA implementation, where the recital on use as a means of exchange is the most likely entry point.
Whether agent-specific licensing emerges. The current design of every major regime treats the agent as transparent, with the principal as the licensed or registered party. There is no signal yet that any major regulator is contemplating agent-specific licensing; the more likely path is principal-side licensing extensions on what authority a principal can grant, what disclosures are required, and what audit trails must be kept.
Whether the existing agency law overlay holds in stress. A hallucinating agent that misroutes funds, an agent acting outside authority on the basis of a model error, an agent that follows a prompt-injection attack into a malicious transfer. The doctrines have answers in principle. The case law is thin. The first major dispute will set patterns for the resource to track.
What insurance products emerge. Errors-and-omissions cover for agent operators, principal-side cover against unauthorised agent acts, smart-contract insurance against logic errors that affect agent flows. Each is being floated by various commercial insurers; the underwriting is early.
How agentic flows interact with significant-instrument thresholds and holder caps under MiCA, the HK Stablecoins Ordinance, and the Japanese PSA. Agentic flows that look like high-frequency person-to-person flows could trigger thresholds in ways the drafters did not contemplate, and the resolution will be jurisdiction-by-jurisdiction.
What to read next
Chapter VIII, 08 wcbdc vs tokenized deposits, is the immediate prerequisite for understanding why agentic commerce lands at the deposit and stablecoin layer rather than the wholesale CBDC layer. Chapter II, Stablecoin types, gives the instrument taxonomy for the regulated stablecoin perimeters that agentic flows operate within. Chapter V, Permissioned blockchains, covers the chain-level mechanics that determine whether a given permissioned ledger accommodates agent-controlled wallets. The institution pages for HKMA, Bank of Japan, and Bank of Korea track the wholesale and retail tokenised-money posture of each major APAC central bank. The project pages for Project Ensemble and the Payment Innovation Project are the most direct current examples of programmable money in regulated production. The themes pages for singapore project guardian and hong kong project ensemble architecture go deeper on the specific pilots that have shaped the agent-compatible tokenised stack.