How does this compare to using something like exe.dev where you get remote isolated sandboxes over SSH, and assume you add in an http credential injection reverse proxy (bring your own)?
I'm building something that will give me peace of mind that my agents will (whether maliciously or stupidly) not negatively read/write the personal data on my machine. To me, exe.dev (with possibly a reverse proxy to preclude the agent from ever having access to my API tokens) is sufficient.
It sounds like your setup works for your case of running your agents in a sandbox. plimsoll is more for running code for agents or MCP servers. It's not for running agents.
A few interesting things:
I did read that exe.dev handles model keys properly (keeps them off the VM) if you use their built in LLM feature.
Hmm, but their Home Assistant integration at first look has no way to set path limits (but I could be wrong), so an agent that is supposed to only read the thermostat can unlock the door.
And ah! Their FAQ mentions using agent forwarding (aka "ssh -A" ). I would avoid that because anything in the VM can use your laptop's SSH keys while you are connected.
Do your agents run or write their own code? Are the agents your own programs or something like Claude Code? If you like, reply with more info on the kind of agent you have and I can look more closely at exe.dev too.
5 comments
[ 0.25 ms ] story [ 2.8 ms ] threadI'm building something that will give me peace of mind that my agents will (whether maliciously or stupidly) not negatively read/write the personal data on my machine. To me, exe.dev (with possibly a reverse proxy to preclude the agent from ever having access to my API tokens) is sufficient.
A few interesting things:
I did read that exe.dev handles model keys properly (keeps them off the VM) if you use their built in LLM feature.
Hmm, but their Home Assistant integration at first look has no way to set path limits (but I could be wrong), so an agent that is supposed to only read the thermostat can unlock the door.
And ah! Their FAQ mentions using agent forwarding (aka "ssh -A" ). I would avoid that because anything in the VM can use your laptop's SSH keys while you are connected.
Do your agents run or write their own code? Are the agents your own programs or something like Claude Code? If you like, reply with more info on the kind of agent you have and I can look more closely at exe.dev too.