Okay, this one is definitely for us nerdy types out there. We’ve all used some kind of AI service in the past, like OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, or maybe even Microsoft’s Copilot. We use these guys on the web, but have you ever used them locally via your command line?
In this NetworkChuck video, he shows us how to use AI directly in the command line. He argues that it’s much faster and allows you to do things like keep the context of your chats locally in a folder and file structure.
Gemini CLI, for example, is free and in his video, he shows you how to install it and walks through the process. It can access files, run scripts, and allows you to have context management right from your terminal. But he doesn’t stop there.
He goes on to tell you about the awesomeness of Claude Code. This version of the AI service uses Anthropic’s Claude and does pretty much everything Gemini does, but it’s a little more versatile because it uses agents.
Agents are basically specialized versions of AI that you create. First, you create the agent by giving it the parameters and characteristics you want it to have. Then, you have it only answer questions it is an expert at. For example, you can designate one as an expert at networking, and then have that agent help you design the best home lab. While another agent that you designed to be a travel agent can focus on creating an itenirary for your weekend trip to New York.
He then kicks it up a notch and shows you how to put all these agents together to work collaboratively, finding the most thorough responses for you. It is a very powerful tool.
If you’re into experimenting and trying new things, especially on the command line, I would definitely give this video a look. He goes through it pretty quickly, but you can pause and examine things more closely, then do your own research to get the specifics down before trying it yourself.
I would caution you that these AI tools ask for a lot of permissions to access things on your computer and on the web, and you do have to give explicit permissions, which is a good thing. Although he shows you how to get around these safety nets, I would think carefully about doing so and know that you’re running it at your own risk at that point.
Just be careful and have fun.