AI agents write code and query APIs all day, but ask one to actually pay for something and the workflow stalls: billing rails still assume a human owns the account. On October 24 and 25 in New York, nonprofit XRPL Commons hands that problem to hundreds of developers at its 36-hour AI & Fintech XRPL Hackathon, with agent payment safety as the flagship track.
Highlights#
- The Agentic Finance track asks teams to build applications where agents manage funds and execute payments while staying inside limits set by their creators. That is the explicit thesis of the event: autonomous spending only works inside guardrails.
- The pitch is infrastructure that already exists. XRPL settles in three to five seconds at fees of a fraction of a cent, and carries protocol-level escrow, multi-signing, and credential checks, so teams assemble policies rather than write contracts from scratch. RLUSD, Ripple's regulated stablecoin, has passed $2 billion in volume.
- Per XRPL Commons, more than 5.6 million x402 transactions have settled on the ledger, and the XRPL AI Starter Kit wires agents into x402 payments in XRP or RLUSD, plus the Machine Payments Protocol co-authored by Stripe and Tempo.
- Over 410 builders had registered at announcement time. More than $40,000 in prizes are on offer, with bounties from Mastercard, Ledger, Quicknode, and AWS, and mentors from Ripple and the XRPL Foundation. No blockchain experience is required; JavaScript and Python are enough.
- Winners get fast-track access to XRPL Commons grants and incubation. The nonprofit has incubated over 90 projects, and its recent Make Waves buildathon produced 56 mainnet launches. The event closes just before Ripple's Swell conference on October 27 to 29.
Why it matters#
Odelia Torteman, Head of Digital Assets at XRPL Commons, frames the requirement plainly: as agents take on financial responsibilities, their mandates must define permissions, spending limits, and accountability. Hackathons are where those mandates get stress-tested. When spending is autonomous, escrow and authorization boundaries stop being features and become the application itself.