Building an Automated Linux Home Server… from a Budget MacBook?

Añadido: 2026-05-06

[00:00:00]This might be the smartest thing I've added to my home setup. I am building a custom Linux home server for my house that requires me to have some knowledge regarding how to set things up properly.

[00:00:11]And I am talking about setting up Casios tail scale as a Docker component.

[00:00:16]Understanding how to take advantage my 3060 with 12 GB of VRAM to run a local LLM. Peril that with maybe an XCloud. I am trying to even run a Windows 11 VM in order to install Da Vinci project server and take advantage of using my Linux computer as my video file storage. But you see, I don't know it all and I also don't want to waste countless hours troubleshooting and wasting time. So, I did something quite different. 20 days ago, I decided to use this laptop in order for me to try and build my Linux server. And if you guys watched our ThinkPad video where I was using that laptop as my dummy in order to replicate a HomeLap server, well, you know that I fought and I worked a lot in order to set it all up properly. With this, however, I don't feel like doing it all over again. And so, I digged all over the internet and I found a tool who are sponsoring this video in order for me to skip those steps. So, let me show you what this is capable of doing cuz it's pretty cool. Listen, after a few several AI agents, I'm now trying Manis AI. I was told that the goal of Manis is to move AI from chatting to actually doing work and I've been playing with it for almost 2 3 months now. It's different from your typical AI where you ask and it responds with text. With Manis, you ask and it creates a plan, executes steps, and it tries to deliver the best results. Sometimes it might even ask you questions. Like, check this out. If I tell Manis that I am building an Ubuntu server, home server, it gives me a list of parts that well I now currently own.

[00:01:53]Manus is intelligent enough to ask questions. What is the primary purpose of the server? What kind of output would you like? What is your experience level with Linux and Ubuntu server? A lot of these questions will serve in order to understand your goal and the level of expertise in this field so it can go ahead and guide you as best as it can.

[00:02:13]But you see, Madness is able to go above and beyond just giving you a step-by-step document. It can actually execute tasks in its own environment.

[00:02:21]Like, check this out. This here is an installation guide website that it created based on my prompts. How freaking cool is that? It's got a really good design to it. Look at that. With proper sections, proper padding. It is gorgeous. It's also got its own navigation menu, so that's cool. And some references. And Manis uses multiple AI sub agents that work together in order to make it all happen. It's got a planner agent that breaks your request into steps. It's got an execution agent which runs tools like browsers or code, a knowledge agent that gathers data and context, and a verification agent who will check results before delivering them. Last year had gone viral because it was one of the first general AI agents capable of operating almost independently. So, Meta decided to to buy them. And if you guys watched the latest video, well, you know that the MacBook Neo is capable of doing quite a lot, especially for people who want to get into development. Now, aside from all the things I mentioned in that video, like learning iOS development, learning web development, trying to learn the terminal, learning about Git, all those cool things, I've been very much using this MacBook to set up this bad boy right here. And instead of using a keyboard and this screen to interact with it, I've been able to pretty much SSH into the server. And yes, I know a lot of this stuff runs on the cloud, but because of it, Manis is capable of doing quite a lot of things. You can upload files, images, and so on. Use different skills to accomplish different tasks.

[00:03:50]You can even add custom skills depending on what you want. You can also create meeting minutes and voice chat. But you can connect apps in order to automate certain workflows. On my end, I connected GitHub. You see, my goal is to build infrastructure as code for my home lab. Meaning that I want a home serverbased setup in my GitHub repo like that. Every new Ubuntu machine I create can pull from that repo and install/configure all the tools I need in the same way.

[00:04:22]And Manis can help out with this if you ask it to. The best approach for me is to give it a very structured prompt. You know like I want to create a complete GitHub repo that acts as a base template for Ubuntu homelap servers. Each server should be able to one install Ubuntu server, two run a bootstrap script, three pull the repo, four automatically install and configure the entire stack.

[00:04:51]And the technologies to use are pretty simple. Use docker for container runtime. Networking wise use tail scale.

[00:05:00]I want my AI stack to be lama running within open web UI. I want automation with N8N some virtualization with KVM by using libvert with key mute and installing Windows 11 and then of course Kaza OS. If you really give it a structure template of what you want, it will go ahead and build it. Minus will use its own environment and terminal to build that for you. And all the agents I mentioned make sure it can deliver the best and most accurate results. based on your prompt. So instead of me manually setting everything up, I'm starting from something that's already been built for me and then bringing that into my own system. Now this example might seem like uh some kind of magical plug and play setup where you install Ubuntu, run one command, and everything just works. But that's not really what's happening. I used an AI agent to build a template for all my servers in its own environment.

[00:05:55]So instead of manually installing Docker, Lama, Tailscale, VMs, and all of my tools every time, I use manage to store those instructions in a GitHub repo. Then when I build a server, it pulls that repo and sets everything up automatically. I still have to configure things like tokens, uh, environment variables or hardware settings, but I only do that once. After that, every new server can rebuild the same environment in minutes. So now I can just pull that repo, run install and my server is basically ready to go. Now I do have this tool called time shift. Time shift is actually cool because it takes a snapshot of the system at a current given time. For example, mine is a fresh installed baseline. And I took a snapshot at exactly that point in order for me to go back in time and keep testing my manus.ai AI Bootstrap file, which is how I've been able to troubleshoot and test things pretty easily. And so with that, I decided to build something on top of it. And that something is this mobile friendly web app for my server. It's pretty much a full home lab control that you can access with the help of your phone.

[00:07:06]Check this out. Okay, I I also had the idea to kind of allow myself to control my Linux servers from my phone. As many of you know, well, I daily and S26 Ultra. I love this phone because it really truly feels like a mini computer.

[00:07:21]But what if I was able to give it some more use? Like use it to connect into my home server for example. What if I could just open an app on my phone and and manage everything like restart containers, check CPU usage, open a terminal? Basically a control center for my servers. Normally building something like this would take a few days. The main reason we're getting this error is because we are using this web app through manus.ai which means that obviously it doesn't have access to our local host and it doesn't have access to tail scale. And so that's why well you can't see any of the telemetry that I kind of want to showcase. But right now I have manis AI trying to deploy this locally so I can actually show this to you. And it's really important to know that you kind of need to know what you're doing in order to properly troubleshoot stuff because you will run into some issues, but the more information you give Managed, the quicker you can troubleshoot and set up things properly. And you'll see that as you're trying to troubleshoot issues and trying to fix your app, Manage really does a good job at deploying agents in order for you to make sure things run properly. On my end, I needed to make sure my web app would work on local host in order for me to connect it to my Linux server, open an SSH terminal, monitor CPU and Docker containers, and run commands. And you can, by the way, specify the stack for these type of projects. Nex.js for the front end, Node.js for the back end, Docker for deployment, and an SSH terminal using Xterm.js.

[00:08:55]But you see, Manis doesn't just spit out code. It actually builds everything step by step in its own environment and it breaks the project into steps, builds the file structure, writes the backend, creates the UI all inside its environment. And this is the result. So check this out. I've got the MacBook Neo running this entire web app locally and I'm using Docker in order to, you know, start up the project. It's really cool because you can clearly see CPU usage, memory usage, disk usage, uptime. You can see so many things within this dashboard. Like take a look at it.

[00:09:33]Pretty freaking cool. Now I am still trying to fix a few minor things and minus is doing that for me. All I need to do is just pull my project, restart Docker, restart that container and everything just works pretty much again.

[00:09:48]And I can just keep troubleshooting and making sure that all the little features I want implemented within this web app get to work at some. I just find it crazy that with this little machine, you can get so much done. And of course, Madness AI has its own dedicated app.

[00:10:04]Actually, you know what? Let me open a new window right here and let me take a look at some of the telemetry for the MacBook with all the things I've got going on. RAM usage is not too bad. CPU usage as well. The fact that this little laptop can do so much is actually really, really impressive. So look, it's really quick at doing stuff. Let me get pull this. Bam. Then I can go ahead and just Docker Compose down.

[00:10:35]It's removing it. Bam. Bam. And then we can go ahead and restart that.

[00:10:41]Look at that.

[00:10:44]Yep. That is pretty cool. And I can just go ahead and once this finishes right here, I can go ahead and restart my app.

[00:10:58]Voila. And now my terminal works cuz that's what I just fixed.

[00:11:03]Awesome. One thing you should know though about Manis is that it doesn't run free. Okay. Uh instead of charging per message like most AI tools do, Manis works on on a credit system. Every time Manis runs an AI agent, it consumes credits. And the bigger the task is, well, the more credits it uses. Like for example, okay, for this server project, I started my journey with 20,000 credits be before trying to fix my web app, right? With a few prompts, like when I started troubleshooting the HomeLab control app in order to connect it to my local network for testing, I consumed about 2,300 credits. What tends to consume a lot of credits, though, will usually be things like building entire apps or running long agent workflows.

[00:11:50]like they've got this new agent feature that allows you to run agents 24/7 if you need to uh on apps like Telegram. I know WhatsApp messenger and line integration is coming soon but uh think of it as a claw built into manage really and I've personally used this a bit for everything. In fact, this past month, I've used it for stock analysis, uh, thumbnail generators and modifiers, trip planner, checking out some real estate, and even transcribing audio in order to help me work faster, which works really well. Personally, Mannis feels like you have an admin working for you at all times. Like, I remember when I was editing a potential thumbnail for one of the vlogs I posted, it was able to cleanly remove the people in the background and regenerate what was needed. And this, by the way, is a skill that you can further enhance within the settings. Manus also has the ability to generate video with the help of some tools like Nano Banana. It's got its own uh MD file that allows for your agent to understand how to do so. Minus is always ready to read and understand your requirements and build you a video based on what you need.

[00:13:00]>> Starts before the sun.

[00:13:08]Connection forged in precision.

[00:13:23]Stock analysis is another skill I've been very much playing with. With the tech sector being all over the place, I like to study companies and their fundamentals in order to understand where I want my money to go. Minus knows to analyze stocks and and companies using financial market data. It actually gets uh company profiles, technical insights, price charts, insider holdings, and SEC filings for comprehensive stock research. It's a very good skill to enable and want to use in this economy. Manis really is like an AI operator that can help you build, research, and prototype ideas all in one place. Now, obviously, this isn't magic. You still need to guide it, write good prompts, and understand what you're building. But once you start thinking in terms of agents that can actually execute tasks instead of just answering questions, you realize this type of AI can actually help you build real things, not just talk about them. And honestly, I think tools like these are going to completely change how people create software, automate work, and experiment with new ideas. I don't know. What do you guys think? Is this something that's interesting? because I I tried to put it into practice and actually build something interesting with it. I personally think it just helps me save a lot of time and create cool things with ease.