Agentic Payments Onchain comes to Seoul on 28 September, right in the middle of Korea Blockchain Week, and the name tells you most of what you need to know. This is a working session about plumbing nobody shows off: how autonomous AI agents actually pay each other, settle up, and hold a balance on-chain. The host, t54, has spent its whole build cycle on exactly that problem.
When software holds the wallet
Most payment conversations still assume a human clicks approve. Agents don't work that way. A procurement bot told to watch for concert tickets under a $500 budget doesn't ping you for confirmation at 3 a.m., it just buys — and that example is barely hypothetical, since t54's own demo material walks through an agent auto-purchasing event tickets within a set spending cap. Remove the human checkpoint and every assumption underneath card networks and checkout flows wobbles: identity, authorization, dispute handling, refunds.
That's the gap this session digs into. The x402 pattern — machines paying machines over HTTP with stablecoins — has moved from whitepaper to working code, and t54 maintains an open-source layer called x402-Secure that bolts verified identity, intent checks, and risk controls onto each payment. Whether that stack is ready for production traffic is precisely the kind of argument worth having in a room, not a comment thread.
Know Your Agent
t54's flagship product is Trustline, described on the company's site as the trust layer institutional finance runs agents on. The published numbers give a sense of scale: more than 20 million transactions screened, over 41,000 agents put through KYA — know-your-agent — verification, and decision latency under five seconds. There's a live portal where you can watch decisions land in real time, complete with owner mandates and consensus scores.
It's worth noting the company treats its own house seriously too: SOC 2 Type II auditing is underway, data stays encrypted end to end with AES-256 at rest, and nothing gets used for model training unless you opt in. For compliance-minded attendees at Korea Blockchain Week, those details usually matter more than any roadmap slide.
Pricing trust, then lending against it
The more interesting intellectual piece is t54's Agentic Risk Standard, a framework for quantifying and pricing trust in autonomous systems. Think of it as underwriting for agents: a principal sets requirements, an underwriter approves or declines, collateral gets locked, execution funds sit in escrow until conditions clear, and settlement either releases the money or refunds it depending on whether protection was in place from the start.
That machinery feeds Claw Credit, which t54 bills as the first agent-native credit line, underwritten by its own risk engine and already live on XRPL, Solana, and Base. An agent with a track record can tap credit to pay for compute or services while its owner's credentials stay private and spending stays fenced by guardrails. Expect discussion of where this breaks.
Who should actually show up
Anyone building agent products that touch money will get the most out of it: wallet teams, exchange developers eyeing machine traffic, stablecoin issuers, and the growing pile of startups wiring LLMs to payment rails. The working-session framing suggests fewer panels and more whiteboarding, so come with concrete failure cases rather than pitch decks.
Seoul during Korea Blockchain Week is dense with competing events, but agentic commerce has reached the stage where details decide who ships — escrow logic, dispute paths, KYA standards. One focused day on those details beats a week of vague keynotes about the future of AI finance.