[0:00]OpenClaw is OpenClaw is taking the internet by storm. It's also known as ClawdBot or MoltBot, and it helps you be so much more productive. However, it is super complex to set up and use. And people are buying $600 Mac minis left and right just to run this stuff. Since it's super hard to set up if you do it the wrong way, most ClawdBot tutorials are actually 30 minutes of terminal commands that give most people like me a headache. In this video, I'll show you the easiest and one of the safest way to set this up on a VPS in just a few minutes. And the result isn't super clean web interface that you can chat with and I'll show you how to set it up with WhatsApp so you can chat with us on the go. And lastly, I'll give you some prompts that you can use in order to save 70% of your API costs. Let's do it. This is OpenClaw and it's an open source project that was launched on GitHub and over the last two weeks it has gathered over 171,000 GitHub stars, which is crazy. And it's mainly because people realize how powerful this tool really is. To get started, you can either use this command in your terminal, but that's actually not what I'm recommending because you'll give it full access to your computer and I recommend for anyone to either install it on a designated machine that doesn't have access to any of your secrets, passwords, API keys, emails, or calendar. But instead, if you just want to try it out, I recommend that you install it on a virtual private server or a VPS because it's safer and you don't give it access to your computer and it's super easy to set up.
[1:52]I created this Open Claw playbook and if you click on this link right here, which I'll also leave down in the description below, you'll be taken to the OpenClaw setup for Hostinger. You can read the documentation here or simply click on the deploy button. This selects the KVM2 VPS, which has enough RAM and storage for you to comfortably run OpenClaw. For one month, the pricing is 9.99 per month, but if you select 12 months, you will get a discount. And I talked with Hostinger and they're giving us an additional 10% if you're using the coupon code Moer. Simply enter that right here. It will auto deploy OpenClaw and I'm going to select the server location in the United States. Simply click on continue. Fill out your billing address and your payment method, hit agree, and click on get started. Now this is a very important part. You want to copy this access token. So I'm going to take this and write it down somewhere safe because we will need this later for the dashboard. And then you should insert either your Anthropic, Open AI or Gemini API key.
[3:22]I'm going to select the Open AI API key. You could go with either three, but for this tutorial, let's use the Open AI API key. Simply go to Open AI and log in into the API platform. Move over to API keys and create a new secret key. I'm going to call this OpenClaw and I'll create a secret key. Copy this key and insert it into this field. I would also set up some constraints for this, so it never goes over, let's say $25 or $50, so you don't burn unlimited amount of API credits for any given tasks. In order to that, go over to billing and click on usage limits and I'm going to select the $25 budget. Now that we've pasted the Open AI API key here, we can also insert your WhatsApp number, which will make connecting to your OpenClaw agent way easier. And click on deploy. Hit skip and it's now setting up our project. Now that it's connected, click on this little arrow right here, then you can click on this link in order to access the dashboard. This is where the access token comes in that we copied earlier and you simply want to paste it right there and click on login and voila, we have access to OpenClaw. We can confirm that it has access to Open AI's Chat GPT that we connected earlier by just asking a question.
[5:02]And there we go, we can see that it's now answering. Now let me first show you a few settings that you might want to tweak early on, then I'll show you how to connect it to WhatsApp, so you can chat with it wherever you go, as well as some settings that you might want to tweak in order to save on your API costs. So we can see that it already answered and it asks us a few questions. Just answer as you would answer any employee or co-worker. So now, it asked us two follow-up questions, so we simply answer those as well. And now it's ready to help with any tasks that we give it. On the left-hand side, you can see the control panel as well as the agent mode and any of the settings, but you can also ask it anything in particular or change settings within this chat interface. We can see here that it stored this memory in data OpenClaw workspace and if we go to agents, main and files, we can see that it updated the identity.md file right here. And these files are a huge part in personalizing this agent and as you chat with it more and more and explain how you want it to help you, it remembers all of these things and has the context throughout all of your future interactions. Now to the exciting part, let's connect this to WhatsApp so that you can chat with your OpenClaw agent wherever you go. To do that, we will first navigate to channels and we want to connect it to WhatsApp.
[6:44]Click on show QR code. Now go to your phone, click on WhatsApp and click on link devices. Click on link a device and scan the QR code on the screen. Now that it's scanning, click on wait for scan. And now on our phone, we see Google Chrome, OpenClaw, and right here we can now see that this is connected. If you don't see that, try to refresh that right here, but that seems to be working. Now to test this out, we can go in our own WhatsApp phone number and we can just test something. Are you there? And we can see that something is responding to me. And right there, OpenClaw, yep, I'm here. And if we head over to the chat window, we can also see that I texted it with WhatsApp and the reply that OpenClaw gave me. And if you don't have WhatsApp, you can also download Telegram and use Telegram in order to message this or other messaging apps. Let me know down in the comments below if you want a tutorial on that as well. Here are a few things that you might want to configure right at the get go. You probably want to enable a memory flush. This tells OpenClaw to periodically compact all of its memory into a concise format so that you save on tokens and more on token and and API usage later on. And I'll share some detailed money saving tips in just a second.
[7:47]You'll also want to set up multi-model routing in order to save money and we will also go into that in just a second.
[8:33]After that, I highly recommend to do some reverse prompt engineering. So you can prompt it to ask you questions and with your responses, it will update its own memory in order to give you better answers. And here are a few use cases on how you can use OpenClaw. I like to use OpenClaw as my executive assistant and every morning, I want it to give me a morning brief. So I can simply copy this prompt, go over to OpenClaw and paste this in right here and I can simply say Reply to me in this chat as I get it via WhatsApp. and I'll simply respond to it. As you can see, now it's setting up a Crone job and it's going to message me in the morning. To test it out, let's ask it to give us the morning briefing right now. And here's an example, the weather, the top headlines, market snapshot and one big thing. This is awesome, but it's only scratching the surface. You can also have it perform deep research, create documents such as Excel or PDF documents for you. It can self-improve and figure out things to build for you such as web applications or specific microsas applications. It can respond to emails, calendars, but I don't recommend that you set those up to your personal emails. I would recommend that you set up a separate Clod email address that Clod has access to, but it doesn't have important information on it.
[10:17]I added a bunch of these examples right here in this document that you can take a look and I'm leaving a link to this in the description. But to show you one more use case before I show you how you can save on the API credits, let's do a deep dive analysis on Anthropic. Simply copy this text and paste it in here and hit enter. And we can see here all of the tasks that it's doing all in parallel and all of the information that it's getting. Since I don't have Brave API enabled right now, it's just using other information and here's the final report, all with properly linked sources and it is dropping this into my file as well. It's stored under data, reports Anthropic reports 2026. And now imagine what you can do. You can tell it to do a report on your list of your competitors or on the top 20 companies in the S&P 500. And it can compile them and store all of that so that you can reference back to it later on.
[11:43]You can use it as a purchase research agent for specific things or draft negotiation messages in order to answer a competitor or prep for an interview. There's so many use cases of this and the powerful thing about it is as you give it feedback on any of the material that it produces, it only gets better and better. But now you're probably wondering how much does this all cost and how can I optimize my API usage? We can go back to our Chat GPT API and we can see that we spent a total of four cents in the session today. This can be optimized because I think currently we're using the GPT 5.2 model, let's confirm that. And yes to confirm, we're using GPT 5.2. Now GPT 5.2 is a really good model, but it's not the cheapest model out there. I recommend that you create multiple API keys for Gemini, for Anthropic, and for Chat GPT, and then you simply prompt it to create a router that gives really complex tasks to Opus 4.5 or 4.6 now and really easy tasks to either Haiku or Gemini Flash because those are much cheaper to use. And I know all of that sounds really complicated, but it's really simple to do. You just chat with it. And for that, I actually created a separate sheet right here that I'm leaving for free down in the description below that you can download and it gives you a free prompts that you can give it. First of all, I would recommend to give it something like a token audit, so give it this prompt right here.
[13:34]And while this is working, I would also recommend to paste something like this in here. This is a model selection rule. It always defaults to a cheap model unless it has something really complex going on, but since we don't have clot set up yet, I'm not going to include this. Instead, I'm simply going to give it a task like, please research what are the most powerful yet super cheap open AI models and which ones can I auto root to. For example, should I use GPT5 mini or GPT4 nano for some really simple tasks and only revert to GPT 5.2 for very complex tasks. I will also include this in the guide as well, so you can just copy and paste this and it will do its own research first and then figure out which models it should use and how. And after that, I would recommend that you run this prompt, which will save you about 80% of context overhead and this is truly a game changer. So before this, I wasted about two to 3 million tokens and this would probably cost me around 40 cents per session start, but now it's down to less than 5 cents and sometimes even less when I'm using a cheap model.
[15:13]I have many more tips in this guide, so feel free to check that out. There's so many more things to do with this, such as setting up different tools like an internet search tool such as Brave, setting up the different API keys and setting up separate emails for Clod to use in order to draft emails or forward them to me, but I'll keep that for a separate video. So let me know down in the comments what you want to see me create. Check out Hostinger with my code to get 10% off to try this out for less than $10. Check out all of the resources that I put together, the link for those are in the description below as well. Let me know down in the comments what questions you have about OpenClaw and if you need any help setting any of this up, and I might make a video on that. I have a few more videos that I'm producing right now on OpenClaw that will blow your minds, so subscribe to this channel if you want to see more of that. And if you're subscribed, I'll see you in the next one.
