I'm a builder at heart, the kind of person who'd rather be in a terminal than a slide deck. A while back I started building AI agents for cloud infrastructure: agents that work on real cloud accounts, real Terraform, real production systems.
The agents themselves turned out to be the easy part. The hard part was everything underneath: keeping them running for hours or days without falling over, recovering when a worker dies mid-task, asking a human before doing anything risky, and showing exactly what an agent did, step by step, down to the token and the dollar.
So we kept fixing the foundation. That took about two years, and it became Opengeni: an open-source platform for running agents you can actually trust in production. It's model-agnostic and self-hostable, and it comes with sandboxes, approvals and full visibility out of the box.
Hey Product Hunt! I'm Davlet from the Opengeni team.
For the past two years we've been running AI agents against real production cloud infrastructure at Cloudgeni. The model was never the hard part. Everything around it was: runs that die when a worker restarts, code that needs somewhere safe to execute, OAuth tokens for every user, tools to wire up, a human sign-off before anything risky, and a clear record of what the agent did and what it cost.
We kept rebuilding that layer, so we turned it into Opengeni.