How to Run A FREE Local Computer Use Agent!

Today, I'm going to show you the new powerful computer use AI agents that you can run locally with Codex and Claude. And you can now run a fully local computer use agent on your machine completely free, no data leaving your computer. And in this video, I want to show you exactly how to set up a local AI agent that controls your computer, opens up apps, writes notes, browses web, or running on your own hardware. I'll also show you how to wire it up with Claude code or Codex.

So it actually does useful work for you in the background while you carry on your day. The key is one tool most people have never heard of. I'll show you exactly how to install it and get it running in minutes. Stick with me to the end because I also show you the one thing you have to set up first or the whole thing falls apart.

And then I'll show you a live demo so you can see your agent working on screen in real time. Let's get into it. So today we're going to be looking at how to use Codex or Claude code and local models with computer use, right? Now, if you want to run local models with Codex or Claude, this is actually a question from Karl.

We've saying like, it's a long subject, but basically what he's trying to do is figure out how to run either Codex or Claude code with a local model, but also run it with computer use, right? And how to use it directly here. So there's two elements to this, right? Number one, if you want to run local models with Claude or with Codex, you can do that via Olama, right?

Olama is the simplest, easiest way to do this, right? So the first step is that you want to download a local model. Now, if you want something that's really powerful, you need a really good setup. And what I mean by that is, for example, you want a DGX Spark or something like that, right?

To run really powerful local models. And that's how you're going to get the most out of it. If you don't have a really good local setup on your computer, even like a Mac studio or Mac mini is okay, but it will just run Jemma for, then you can use one of the cloud APIs, right? So that's step number one, is make sure you have a good setup and then download a local model from Olama, right?

Now, once you've done that, so for example, if you wanted to download Quen 3.6 plus, you can just make sure you have Olama downloaded. Once you have Olama downloaded, then you are going to copy this into your terminal, right? So we've got Olama over here and we would just go to terminal, paste that in, and that will actually pull in and download the model. I'm going to stop it because it would take quite a while and it takes up a lot of memory, but that's basically step number one.

So you download Olama, you pull in the local model, as you can see right here. Once you've done that, you can then set up, for example, like Cloud Code or for example, Codex using these local models, right? So for example, let's say we wanted to use a local model with something like Cloud Code, right? We could, I'm going to do it with a cloud model, but it's the same process with a local model.

So let's say, for example, I wanted to run it with DeepSeq, right? I just copy and paste a command like this and it pulls in Cloud Code, but it's running with DeepSeq, right? That's the biggest difference. We can do the same thing with Codex.

So if we paste this in, we can now run DeepSeq with Codex, right? And so you can run these local models with Codex directly there, right? That's how you can use these models. Now, if you want to use computer use, you would just give it the details on exactly how to use that, right?

So there's quite a few different, because bear in mind, you're running this inside the terminal, right? If you're on the cloud desktop, it's a totally different scenario. But if you're running this inside, for example, Cloud Code with a local model, then you would give it the instructions to use computer use. So I'm just going to check, can you use computer use inside Cloud Code with DeepSeq v4 flash?

And we'll try the same thing inside Codex as well. All right. Now, what we need to do from here is we need to use something called CUA driver, right? Which is the skill that can open and use computer use, right?

So for example, here, I want to say, use CUA driver skill to open up Google notes, sorry, to open up notes application. So you can see, I don't have this open yet, but we're going to use CUA driver, which is a skill for computer use with this model to open this up. Then we just need to allow the settings with CUA driver, which I'm doing here, right? So we've just allowed that.

And now it's opened up notes, as you can see. Let me try another example here. So if we go inside here and we're like, okay, write a note about today inside notes, and I'm just going to close notes again. And we'll say using CUA driver.

And the cool thing about using computer use with CUA driver as a skill is that it's running in the background and it's working without me, right? So I can carry on working on Cloud inside the tunnel. I can continue doing whatever I'm doing without getting interrupted by the computer use, right? So it's pretty powerful here.

Now, honestly, is computer use amazing on any sort of platform? I don't think so, but it is quite, it is useful to run in the background and do other stuff. It can get stuff done. Hermes can do the same thing.

So if we go inside Hermes here using CUA and we say, okay, use computer use, write a blog about Hermes inside notes application. It's going to prepare computer use as you can see. And now it's opening up notes. So this is notes over here.

And you can see it's navigating on the screen and starting to write on what it did, right? So this is May the 14th, started working with CUA driver today to automate macOS apps from the terminal, right? Notes can be created and they'd say, blah, blah, blah. Here's a key takeaways, right?

And so you can see here that it created a full note using Olama with a deep seek and CUA driver, right? So you can create notes. Is it smooth? Is it easy to use?

No. But if you're insistent on using computer use, that's how you can do it. And that's how it works step by step. Right?

Pretty cool. So just to recap on the step set, you make sure you have a good local setup. Then you're going to download Olama. Then you're going to download local model.

Then you're going to run core code using Olama, set up CUA driver and use CUA driver for computer use. That's basically it. That's the whole process. So if you want to get a full training on how to use computer use, we actually have a full section on using it with Hermes agent, which we created yesterday.

You can see that on Hermes computer use here. It's a full video tutorial and step-by-step guide. As you can see, and that shows you exactly how to use computer use step-by-step using these AI agents. And if you want me to ask, answer questions like we did today for Kyle on a video, then feel free to join the Aircraft for Boarding.

You can post inside the community, answer questions like this every day. You can ask questions, get help and support whenever you need to. You can get new trainings inside the classroom on all my new stuff. You can see all my new daily updates on the left hand side here.

We also do four weekly coaching calls where you can jump on live calls, share your screen, ask questions about your setup, et cetera. And then inside the map, you can connect with people in your local city who are using AI agents like you. So feel free to check that out. Thanks for watching.

I'll see you on the next one. Cheers. Bye bye.