SUPA Blog
Field notes on running a principal cross-border business — payments, foreign exchange, signed documents and operating discipline.
Plainly written. No fluff. Articles on payments, foreign exchange, registration discipline, signed documents and the operating decisions a principal operator actually has to make.
Multi-Agent Treasury: When Your Finance Team Is Software
What it looks like when agents handle payables, receivables and FX under one policy and one ledger.
Read →The Machine-Speed Ledger Problem
Why daily settlement cycles and fees designed for humans cannot serve a constant stream of small machine-to-machine payments.
Read →Verifiable Intent: Proving an Agent Was Allowed to Pay
Explains the mandate-proof layer that lets a bank confirm an agent's authority before releasing funds.
Read →x402, ACP, AP2 and MPP: Mapping the Agent Payment Protocols
A comparison of the emerging checkout, mandate, card and machine-payment standards — and where a settlement layer fits beneath them.
Read →Agent-Assisted Commerce: Connecting the Purchase to Its Authority
How a purchase carries verifiable proof of who authorised it, from checkout to settlement.
Read →Escalation by Design: When an Agent Should Ask a Human
How an over-limit operation becomes an approval request — with the invoice and counterparty attached — instead of a hard rejection.
Read →Prompt Injection and the Money Layer: What Agents in Finance Must Survive
A threat-model piece on prompt injection and manipulation when the agent in question can actually move funds.
Read →Agent Identity: Why an Agent Should Never Be Anonymous or Own Anything
The design rule behind revocable, attributable, property-less agents — and why anonymity is incompatible with regulated money.
Read →Mandates Below the Model: Why Agent Limits Must Be Unarguable
Explains why policy has to sit beneath the model — where an agent cannot reason, persuade or prompt its way around a limit.
Read →The Agent Financial Passport, Explained
How a verifiable passport carrying principal, mandate, limits, permitted countries, categories and a policy hash gives banks something concrete to check.
Read →Who Pays When an AI Agent Gets It Wrong?
A framework for liability across the principal, the agent platform, the model developer and the payment provider when an agent exceeds its mandate.
Read →Know-Your-Agent: The Next Chapter After Know-Your-Customer
Why regulators and the IMF are asking whether an identity framework built for humans can stretch to machines that transact on their own.
Read →