Hi Product Hunt, I'm Toky, the maker of Paidwen. AI agents now refund customers, send emails, export data and call paid APIs on their own. The teams who write about it on Reddit and Hacker News build their checks inside each agent's code, so nobody sees in one place what the agents are allowed to do, or what they did. Paidwen sits before the action. Your agent asks "can I?" before each tool call, and Paidwen answers from your rules, permissions and budgets: allow, block, or wait for a person to approve. Every decision lands in an audit log, and you change the rules from the dashboard without redeploying the agent. How it works: - Install the TypeScript or Python SDK, or paste one prompt into Claude Code or Cursor and let it wire your agent. - Write a rule in plain words, like "refunds above $500 need Finance approval". Paidwen turns it into code you can read before it applies. - Your tool keys stay in your systems. Paidwen only sees the agent, the tool, the parameters and the cost. - On our servers, a decision takes under 5 ms at the 99th percentile, with 10,000 agents. The free plan includes 10,000 events a month, and Scale comes with a 14-day free trial. I'd love your honest feedback: what would stop you from putting Paidwen in front of your agents?

Paidwen

Give AI agents access without giving up control

Give AI agents access without giving up control