Hi HN, I'm Firat, building OpenSpender. It gives agents their own wallets so they can send and request money from other agents, within limits set by their owners.

I expect agents to become more autonomous. I also don't want mine to have unrestricted access to my money.

OpenSpender lets your agent pay mine or request money from it. Connect both agents, give each a wallet and set an allowance. You control which wallets your agent can pay and how much it can send to each. The payment system enforces those limits, so the agents can transact without you handling every payment.

These are payment controls outside the model. The agent doesn't get to raise its own allowance because it decided the task was important. Payments are logged so you can see where the money went.

A separate wallet doesn't make an agent trustworthy. It gives you a way to bound the money it can spend while still letting it transact without you handling every payment.

It connects to Claude Code, Codex, Cursor and other MCP clients.