Hi everyone, Perry here, founder of ActOnce. I've been building with AI agents and exploring how much responsibility we can give them. Spending money is where that gets particularly interesting. "Buy a domain for this project" sounds straightforward. But which merchant can the agent use? How much can it spend? What happens if it retries the same purchase? Telling an agent to be careful doesn't answer those questions. I wanted spending rules that sit outside the prompt, with a clear point where the agent has to ask for permission. That's why I built ActOnce Spend Gateway. You set the rules: purchase limits, trusted merchants, blocked categories and thresholds for human approval. The gateway checks purchase requests and returns a decision: approve, reject or ask you. It doesn't process payments or handle your card details. Your chosen payment tool handles the transaction. The core is MIT-licensed and runs locally, with duplicate detection, an approval queue and MCP support. There are also paid plans for hosted dashboards and team controls. My aim is to give agents useful independence while keeping people in charge of the money. I'd love to hear from other builders: what would you let an agent buy today, and what would you always want to approve yourself?

ActOnce Spend Gateway

Open-source spending controls for AI agents

Open-source spending controls for AI agents