Toad
The whole pitch is one sentence, "Point your agent at any machine you own," and honestly that's enough to get me interested. Most agent tools keep you locked to a single box or a cloud sandbox. Toad is betting that the real value of an agent shows up when it can reach your whole fleet, your home server, your work laptop, that old Raspberry Pi collecting dust.
That's a real shift, but it puts a target on your back. Giving an agent broad access means you're trusting both the agent and whatever model is driving it. I'd want to know how Toad handles permissions, whether it's SSH keys or a daemon, and what happens when an agent makes a mistake on a machine you actually care about. The demo probably looks great. The recovery story is what I'd press on.
Still, I think this is the direction things are heading. Agents that can only chat are toys. Agents that can touch your machines are tools. Toad is early and thin on details, but it's pointing at the right problem.