Ceadly: Human approval for AI agents, sealed as proof

Hey Product Hunt 👋 Three weeks ago, Ziya and I left our jobs to build this full-time. The problem: AI agents don't just answer questions anymore — they send emails, issue refunds, write to production databases. When one misfires, the incident log shows an agent name and a timestamp. No approver. No proof anyone was watching before it happened. Ceadly fixes that with one line of code. A decorator (@guard) pauses any consequential agent action before it executes. A named reviewer sees exactly what's about to happen and picks: Approve, Reject, Modify, or Request More Info. The moment they decide, it's hashed and timestamped — tamper-evident proof of who approved what, and when. We're not another observability dashboard telling you what already happened. We're the checkpoint that stops it before it does. Still early — three weeks into building the product itself, MVP live, no signed customers yet. We'd love your honest feedback, especially if you're running (or worried about running) AI agents in production. What would make you trust a system like this enough to put it in front of your own agents? FOR QUESTIONS - hi@ceadly.me