Best Practices

[chrismuzyn]

Hi,

I'm completely new to the ai agent thing. I skipped openclaw. I'm familiar with ai coding tools like claude code and opencode. I definitely see the potential of ai agents, I just need to figure out the best way to do it.

I have docker compose upped the production profile, and have integrated discord, and all my models. This leaves me with a single node that has a bare minimum unix environment, almost no tools preinstalled. Adding ssh skills, for example, doesn't usually work because ssh-keygen isn't here, sshpass, etc. I'd like to add an imessage mcp server from a mac, but, apart from adding everything to a tailnet and locking it down, I don't feel good about having that on the network.

These two experiences lead me to think that I should be adding task specific nodes. Maybe, this mac node has my imessage mcp server, this linux node has my ssh keys, this linux node has my ansible playbooks. Is this a good way of doing things, or am I walking down the wrong path?

[eous]

If you do curl -fsSL https://raw.githubusercontent.com/turnstonelabs/turnstone/main/run.sh | bash it will walk you through the basic setup of an up to ten node (think concurrent workers) stack with postgres + caddy self-signed tls http 2 front end proxy.