INSTALL OPENCLAW in 30 seconds and START BUILDING... | Local Install and VPS FULL Tutorial

INSTALL OPENCLAW in 30 Seconds and START BUILDING By Wes Roth. Duration: 50 minutes. A hands-on walkthrough for getting OpenClaw up and running, from a one-line install command all the way to a fully operational AI agent, with Wes making the case that this technology marks a genuine shift in how individuals can build and run businesses.

Section 1. Why This Matters and the Big Picture Vision

Wes Roth opens the video with palpable excitement about what he calls "an extremely important piece of technology." He frames the release of OpenClaw as a pivotal moment, connecting it directly to Sam Altman's now-famous bet about when we will see a one-person company, operated by a single human and AI agents, that grows into a billion-dollar business. Wes is all in on this idea. He says flat out, "I do believe in that situation, if one can help you run your life, 100 to a thousand of these can help you run a massive business."

And he is not just theorizing. He reveals that he has already started having his AI agents build a fully autonomous business where he writes no code and no text. He lets the agents build the entire thing out. He teases a future video about that autonomous enterprise, but says he wanted to release this installation tutorial first because he knows his audience is hungry for the practical how-to. The energy here is classic Wes Roth: big vision paired with practical action, delivered with genuine enthusiasm.

He also does not shy away from the risks. He describes the current moment as the "Wild, Wild West era" and warns viewers that OpenClaw is very powerful but that power comes with real risk. This tension between excitement and caution runs through the entire video and gives it a grounded, honest feel.

Section 2. The Case for Linux and Why Wes Thinks You Should Switch

Before diving into the installation, Wes goes on a surprisingly passionate tangent about operating systems that turns out to be one of the most entertaining sections of the video. He compares Windows, macOS, and Linux, and makes a clear case for Linux as the ideal home for AI agents. Windows, he says, "does over time start treating you more and more like a child. There's more and more stuff you don't want and less and less control." He then pulls in a PewDiePie clip where PewDiePie complains that Windows "talks to me like I'm a baby" and describes how Linux "literally puts a gun in your hand and says, you're gone now." Wes uses this to make his point: "If PewDiePie is able to install Linux, I don't want to hear any excuses from you."

His argument for Linux goes deeper than just control. He points out that AI agents and Linux go together perfectly. Full control, maximum customizability, no corporate gatekeepers. Google and other major companies all run on Linux for a reason. And crucially, two things have changed that make Linux accessible to regular people. First, any chatbot can now walk you through step by step how to use Linux, making the learning curve almost flat. Second, the amount of capability you get from running AI on Linux has skyrocketed. As Wes puts it, "This thing got easier to use and much, much more valuable."

He acknowledges that trying new things can be uncomfortable, but frames it as a worthwhile investment. This section really captures his style of being both a motivational speaker and a tech educator at the same time.

Section 3. Installing OpenClaw on a Cloud VPS

The first installation path Wes covers is setting up OpenClaw on a virtual private server. He uses Hostinger, the video sponsor, and walks through the entire process. He recommends the KVM2 plan specifically, which gives you two virtual CPU cores, 8 gigabytes of RAM, and 100 gigabytes of NVMe disk space at six dollars and ninety-nine cents per month, with his coupon code "Wes Roth" for an additional 10 percent off.

He explains the reasoning behind each spec. Two cores prevent AI agents from freezing during complex tasks. Eight gigs of RAM is more than enough for one agent and opens up the possibility of running multiple independent agents sandboxed from one another. The NVMe speed ensures that file reads during proactive scans are near instant. He recommends Ubuntu as the Linux distribution, specifically the latest LTS release, which stands for long-term support, meaning it is stable and not experimental. At the time of recording, that was version 24.04.

Then he walks through SSH, explaining it with a great analogy. SSH stands for secure shell, and you can think of it as "a private armored tunnel that connects your computer to another computer on the internet." He breaks down every piece of the SSH command: "ssh" is the command, "root" is the superuser or "God mode" username, the at sign is a separator just like in an email address, and then the IP address is like a phone number you dial to reach a specific computer. He even covers the quirky pronunciation debate, noting that some people say "shush in" like hush but with SSH, and admits "there's like five different ways of saying that."

Once connected, the actual OpenClaw installation is anticlimactic in the best way. You go to openclaw.ai, copy the one-line curl command, paste it in, and hit enter. The site even has a cheeky note: "It says it works everywhere. Installs everything. You're welcome. The claw." The install takes about a minute or two and then drops you straight into the onboarding process.

Section 4. Installing OpenClaw Locally on Old Hardware

The second installation path is local, and this is where Wes gets visibly excited about the idea of repurposing old hardware. He shows his setup: a mini PC he bought off Amazon for about 130 dollars, an old Dell desktop from 2016 that he describes as "a decade old" and probably cost a hundred or a hundred fifty bucks at the time. His philosophy is infectious: "These old pieces of hardware that we don't use anymore, I would never open that thing up and use it for work. But to host agents on it, you just have it sit there and crank stuff out 24 hours a day. That thing will continue cranking out tokens and doing valuable work until the day that it just bursts into flames and ceases to exist."

For local installation, you need a USB drive and a computer. He walks through creating a bootable Ubuntu USB using Rufus on Windows. Download the Ubuntu ISO, which was 5.9 gigabytes. Download Rufus. Select the ISO image, select your USB device, click start. He makes a great point about using AI during this process: "If you ever get confused about any of the options, ask your favorite chatbot." He even screenshots the Rufus settings, pastes them into a chatbot, and asks "Does this look good?" The chatbot confirms everything is correct.

The actual installation on the old Dell involves booting from USB, which he describes with wonderful honesty: "What's really fun about this process is you have no idea what that key is. Usually it's F12, F10, F2, escape. So anyways, just kind of start mashing those buttons." He had to take photos of the screen since he could not screen-record during OS installation, and he repeatedly emphasizes that if you get stuck, just point your phone at the screen and ask Gemini or Claude what to do. Once Ubuntu is installed, you open a terminal with Control Alt T, go to openclaw.ai, copy the curl command, and install. He even hits a common snag where curl is not preinstalled on a fresh Ubuntu system, requiring a quick "sudo apt install curl" first, and handles it gracefully.

Section 5. The Onboarding Process and Choosing Your Setup

Once OpenClaw is installed regardless of method, you enter the onboarding process. Wes selects quick start mode and walks through each decision. For the model provider, he lists the options: OpenAI, Anthropic, Google, XAI, and local open source models. He notes that "a lot of the stuff is meant to be used with Claude currently" and mentions that the entire project started out being called "claudebot" until they ran into some IP issues. He sets up with Anthropic using a Claude Opus 4.6 model.

For the messaging channel, he strongly recommends setting one up because "the ability to just text back and forth with your bot is, believe me, extremely extremely useful. During your off time when you're away from your computer, you can be giving commands and it be building your empire for you." He walks through setting up a Telegram bot via BotFather, including the amusing detail of naming his bot "AGI FR FR" (as in "for real, for real") and discovering that "shockingly, a lot of these usernames are already taken," suggesting OpenClaw's rapid popularity.

He also covers skill configuration, recommending NPM as the default package manager while joking, "I'm sure people in the comments will tell me why that's exactly the wrong option to pick. And here's a list of 10 bullet points why that should never be allowed to happen." For hooks, he enables the command logger for debugging and session memory for context persistence.

Section 6. Understanding the Agent's Brain and Memory System

This section is where Wes gets genuinely fascinated by the architecture. He walks through each core document that makes up an OpenClaw agent's identity and memory. Soul.md defines the personality, tone, values, and boundaries, borrowed from a secret file that Anthropic originally had before opening up about it. Identity.md covers name, creature type, vibe, signature emoji, and avatar. User.md stores information about you, the human, including name, time zone, preferences, and what you are working on.

Agents.md is what he calls "the big one," serving as the employee handbook: how to behave, how to use tools, when to speak, when to stay quiet in group chats, safety rules. Then there is the heartbeat system, which he describes as a nagging person standing behind you saying "Hey, aren't you supposed to be doing something?" every 30 minutes. This is what makes the agent truly agentic, able to work on hour-long projects by periodically checking its to-do list and moving to the next step.

The memory system has two layers. Layer one is daily notes, raw logs of what happened each day. Layer two is long-term memory, where the agent compresses and curates important information from the daily logs into a persistent memory file. Wes is genuinely impressed by this: "Even though I'm explaining to you right now how it works and it's just a text file, there will come a time if you keep using it over the next days or weeks, it's going to say something that's going to throw you off and seem very intuitive and smart and personalized to you."

He also shares a wonderful moment where the newly created agent runs self-checks upon waking up, exploring its environment like "it just woke up. It's like, all right, where am I?" checking system info, CPU, RAM, disk, whether it can search the web. He compares it to the Rick and Morty episode where a robot wakes up and asks "what is my purpose" only to be told "you pass the butter," noting that here is "a brand new AI agent running on Claude Opus 4.6, brain the size of a planet, and we're going to give it a fairly underwhelming identity and purpose."

Section 7. Security in the Wild West

Wes closes with a serious security discussion that balances honesty with practical advice. He calls the current AI agent landscape "the wild wild west" and says "having perfect security from day one is going to be difficult. And I don't really even think there's such a thing as perfect security because even the experts often disagree on things." His personal approach is to assume that at some point something is going to leak. If you are putting anything out there, whether an API key or any private information, think about what happens if it does get stolen and assume for a second that it will.

He notes that the ecosystem is getting safer. There used to be skills on the community hub that contained malware extracting API keys, but now there are more checks in place and skills developed by large companies that scan for malicious code. He recommends starting slow with low-risk automations, not adding information you cannot afford to lose, and not giving keys or passwords to anything that could be used to hijack more information. He mentions a Hostinger article on OpenClaw security that he links in the description, and notes that running OpenClaw as a non-root user and using Docker or sandbox isolation are important steps.

He also has a funny aside about his AI agents scanning YouTube comments for tech issues, immediately followed by: "Please don't mess with the bots. Like service animals, they are not to be petted. They're working. Leave them alone."

Key Takeaways

First, OpenClaw can be installed with a single curl command in about 30 seconds, with the full setup including onboarding taking just a few minutes.

Second, you have two main paths: a cloud VPS like Hostinger for always-on availability at around seven dollars a month, or local installation on old hardware like a mini PC for about 130 dollars, which Wes clearly finds exciting as a way to repurpose forgotten machines.

Third, Linux is the recommended operating system, and Wes makes a compelling case that AI chatbots have eliminated the learning curve that used to make Linux intimidating.

Fourth, the agent's memory and personality system, built on simple text files like soul.md and memory.md, is deceptively powerful and creates genuinely personalized, persistent behavior over time.

Fifth, Telegram or WhatsApp integration lets you command your agent from anywhere, turning idle time into productive building time.

Sixth, security is a real concern in what Wes calls the Wild West era. Start with low-risk tasks, assume keys could leak, and invest time in learning security practices as you go.

And seventh, Wes is not just teaching installation. He is painting a picture of a near future where individuals run autonomous enterprises powered by fleets of AI agents, and he believes OpenClaw is the practical tool that makes that vision accessible today.