A-CARD gives AI agents their own virtual cards, minted in seconds from a prepaid wallet, hard-capped to a budget, a merchant category, or a single use. What's different: limits are enforced at the authorization layer in real time, not logged after the fact. Anything unusual routes to a human for approval before it settles not after money's gone. Full transaction audit trail, multi-currency (ZAR + USD), built for South Africa and US markets
hey everyone I'm Tebza, solo founder of A-CARD, building from Johannesburg.
I built this because I kept running into the same problem: AI agents are already buying things API credits, SaaS subscriptions, food, flights but nobody's built payment infrastructure that actually treats them like a different kind of spender than a human with a wallet. Today it's either "hand the agent your card and hope" or "make a human approve every single transaction," and neither of those actually works once an agent is doing real volume.
A-CARD gives an agent its own virtual card minted in seconds, hard-capped to a budget or a single use, with anything unusual routing to a human before it settles, not after. Full audit trail on every transaction. Built ZAR-first for South Africa, with US support from day one, because most of what's out there right now is built assuming US rails and doesn't think about emerging markets at all.
Where I'm at honestly: the platform is fully built and tested authorization engine, ledger, enterprise policy layer, approvals running against a sandbox issuer while I close the live card-issuing partnership. So this is very much a "come kick the tires and tell me what's broken" launch, not a "everything's finished" launch.
If you're building agents that need to spend money or you've hit this exact wall yourself I'd genuinely love your feedback. What would make you trust an AI agent with a card? What would make you not?
I'll be in the comments all day. Ask me anything happy to talk architecture, why South Africa first, or why I think agents need credit, not just card limits.