I think AI infrastructure is still lagging behind the models themselves.

Most tools focus on helping developers write code, but there's much less work around running agents reliably on real infrastructure.

So my personal take is Corv: an SSH client for AI agents (and humans)

It lets agents connect by name, keeps credentials in a local encrypted vault, returns structured JSON, reuses authenticated SSH connections, and handles long-running jobs without relying on tmux or nohup.

It's also a normal SSH client for humans, with an interactive TUI, connection manager, SSH config import, ProxyJump support, ecc..

This is v1.0, and while It's been tested extensively, there are undoubtedly edge cases that haven't been encountered yet. Please use it responsibly. Feedback, bug reports, and contributions are all welcome. Enjoy! More infos on X -> https://x.com/khalid64390