Depneds on your main computer. If you're using linux, then create a dedicated rootless user for your agent. However, ota probably best practice to use a separate machine entirely. This is how I use openclaw

clawi.ai 🤘

Dude, can’t you literally ask the AI how to manage the AI?

How's Bitcoin deas

I have a proxmox server so just slap it on a LXC container and firewall it off from everything else on the network (apart from gateway access). Really depends on the Os and network you have.

I run Hermes on my Umbrel. Super easy to install og decent security, as it runs in a sandboxed environment. I plan on migrating hermes to a dedicated machine though, as I am a little paranoid about the security aspect. But its an easy and reasonable secure way to start playing if you already have an...

Using docker is a pain. I use a vm for hermes and the like.

I’ve been using a virtual private server which costs me about $8/mo for a middle of the road package. I used it for a few open claw instances and now I’m running two Hermes instances in separate containers. The downside is obviously it’s less private because it’s on someone else’s machine ...

Run it in a VM. Don't use containers. Do you want to run a local model? If yes then you can put the model itself in a container for speed. If you want to use an external AI api, then this is all not worth it, as you will leak all that data to an external company. (And yes AIs do steal your data and ...

I happened to do that just yesterday. 'lima' works quite fast to create a VM on any Linux/Mac (windows idk) Then it's the hermes oneliner command and voila. You're set. Lima install is 1-3 liner command + a couple commands to create the VM: https://lima-vm.io/docs/installation/

I went old school, I fished out an old Linux laptop scrubbed it, reinstalled fresh new Linux and installed Hermes and chat to it via telegram and git backed obsidian vaults . created its own ssh keys and gave it controlled access to network etc. I only did that because I wanted to share the PC with...

Im using a vm

1. VM 2. Another PC but set ufw rules so it does not connect to your LAN 3. Main computer: you could do another user but it is not going to be secure. Nobody knows what kind of code the LLMs will generate. Everything is random (temperature setting in LLMs)