5 Tips to Make OpenClaw 10X Better (EASY)

OpenClaw is the most powerful AI assistant that exists right now. But if you haven't done these five things after installing it, you're really only going to be utilizing maybe just 10% of its powers. Most people are going to install this on their VPS or their Mac Mini.

They're going to connect it to Telegram or WhatsApp, send a few messages backwards and forwards and be like, "This just feels like I'm chatting to Claude. " Which is kind of true. Now, the difference between Open Claw being just a chatbot and an actual AI employee comes down to what I'm about to show you in this video.

And if you stick to the end, I'm going to show you a way to save 70% on your API costs when using OpenClaw. So, with that being said, let's get over to my laptop and get started. So, before we get into the meat of today's video, you will need an OpenClaw installation.

Now, there are different ways you can go about doing this. Some people spend hundreds of dollars on a brand new Mac Mini. Other people just go the VPS route.

And for speed and lower cost and just ease of access, you can go ahead and get a VPS here with Hostinger starting at just $4. 99 a month for their basic plan. I use a KVM2.

It's perfectly capable for my use cases right now, and it's only $6. 99 a month. And that gives you two CPU cores, 8 gigs of RAM, a bunch of disk space, and if you want a bit more higher performance, you can go up to the KVMA, and it's only 20 bucks a month.

So, thanks for Hostinger for sponsoring this video. And the cool thing about Hostinger is they have a one-click installation. So, once you sign up, you can just come into your account here.

And what I do is I came into the projects and go into the catalog, sorry. And you can just search open claw and you can come in here and you can deploy and obviously put in your API tokens and you can deploy this in literally one click. Now the caveat is this is not going to be fully secure and that's where the tutorial comes in that I put together on my channel on how to set up openclaw on a VPS securely.

So that video will be linked somewhere on the screen and definitely in the description if you want to go through that process to secure your connection which I think is absolutely compulsory. I wouldn't install Open Chlorin VPS without going through that training. Anyhow, let's crack on with the video.

Thanks again for Hostinger for sponsoring this video. And if you use my discount code growth lab, you're going to get yourself a nice 10% off. So with that being said, let's get into the first thing that you can do with open claw and that is to understand the MD files that you have access to that gives context to your various different bots that you can create here.

It's also going to give it a personality. It's going to give it memory. So I'm going to look at all of the files that you can edit here in Quadbot.

So, one of the things that you can do to check out those files, you can just ask. I mean, I'm going to voice note in here because I use Whisper Flow. What MD files you have access to, including the agents MD, the soul MD, and all of those.

Can you just list those for me and show me what's inside of those documents? So, if you've not created any of those, there probably will be some templates set up, but to give your individual agents a role, essentially, you're going to need to fill out these documents. Now, it's important to figure out and think for yourself what is the goal of my agents because I have three here.

I've got an image creator, I've got a content writer, and then I've got master orchestrator is what we've got here in this chat. So, once it's loaded everything here, the cool thing is it's going to tell you exactly everything you need to know about these files and how to use them, how to edit them, and how to update them. And you can literally do everything from this chat, which is the best thing.

Now, if you want to access your files like in a Finder window, for example, like I have here, I've installed this tool called Mountain Duck. Now, maybe some of you will have heard of Cyberduck. It's a spin-off company from that that allows you to connect to remote servers, and you can ask Claude.

That's what I did. Um, you will need to connect your Tailscale account and authenticate the connection that you're going to make with Mountain Duck. But once you've done that, you'll have the IP address here and you'll just connect using I mean I I'll just show you real quick.

So when you open Mountain Duck, you'll have options to open a connection and what you'll simply need to do is head over to the SFTTP and you pop in the URL that you've got from Tailcale that you've authorized and then the folder path. What I found I've just copied I did this var. Docker volumes.

If you've installed it through Docker with the VPS with the method that I showed you, then you'll just pop that in. You'll put root and then you'll click connect. If you're on Mac, it will ask you to go to your system settings to authenticate.

And then you'll have access to all of the files. As you can see here, there's the config and then there's the workspace. And if I drop these down, you're going to see all the different files inside of here.

So yeah, you can see my agents. We've got these clawbot JSONs. We've got the sub agents.

Okay. Okay. So when I SSH using Mountain Duck into this folder that I showed you on Claude the uh just this path here when I do that what I'm able to see is these two files here.

We've got the config and the workspace. If you just open these go in and you can see got the agents. If I click into here does take a second for everything to load.

You can see in my agents folder I've got the content writer. I've got the image creator and I've got the main. And if we drop these down, you can see models, some sessions, image.

You can explore these these files and see what's in here. These are not the ones that I wanted to show you into here. We can see we've got the agents MD file.

We've got the heartbeat MD file, the identity MD file, memory, soul, tools, and user. So these are all files that are going to give your agent context. So in my agents.

mmd file when you're in a chat you can ask claot to add information to this file and this is essentially a process you know every session before doing anything read the working memory read the soul. md read the user. md all of this stuff right so you can tell it and give it instructions of what to add to your agents.

md file this is kind of like the master file that gives a process and some structure to your agents you can also add in your GitHub repo and things like this Right? A memory. md file.

It's going to give your agent some memory about the things that you're working on. Soul, what it says on the tin. It's giving your agent some kind of character, some kind of flavor, some personality, the vibe, the boundaries, the core truths.

Okay? And this is something that you'll update as you move forward and create some different use cases and agents tools, what it's kind of got access to, and some of the files and the user MD file. This is all the information that it stores about me, the human that it's working with.

So, if you don't have access to Mountain Duck and you just want to do everything here in the chat, it's totally fine. And you can see here from my previous request earlier in the video, you see got them agents MD file. This is essentially workplace instructions, how the agent operates, how the heartbeat behaves, etc.

, etc. Memory MD. This is longerterm memory.

Like I mentioned, we got some group chats. So to add something to memory, you want to say remember this and it's going to update one of the relevant files here. It does also have files for different days, which is super useful.

So the skills that you install are going to be the tools that your agent has access to. We're going to get into that in just a second because this is where things get really, really powerful. Now, I don't have heartbeat.

MD set up. This is basically having your agent checking in on you. You know, some people do it every half an hour, every hour, this kind of thing.

So, if you have stuff that you're working on that you can just set off and work along with your agents in the background, this is great for that. But if you have like one-off tasks that you just do and it gets completed and that's it, then they might not be super useful for you in this case. So, during your heartbeats, there is memory maintenance.

So, it's going to read through the recent memory files, identify any significant events, lessons, or insights worth keeping, and then it's going to update your actual memory file with those distilled learnings. The memory from your day-to-day is stored in a different file. Just to be clear, the soul file here is just the personality and the core principles.

You can add some real character here if you want to go for that. Identity. md is the who I am.

So the name, the vibe, and the role. So this is going to be different for each of your uh sessions here. And on that note, like how do you create different agents?

And it comes down to your sessions, right? And you can actually create a new agent just by talking to your main agent in the chat. It's that simple.

Now, the Bootstrap MD, I've not used this. It's missing. It says here, "This would be my birth certificate for the first run setup.

I didn't go through any setup process here. I just went through the security that I've done and shared in my previous video. " So, that's the basics of the files that your agent has access to.

So, an easy way to get started on editing these is just to ask your agent to interview you based on the tasks you want to complete, the kind of personality you want to give it. Just ask it to interview you. so that you can fill out these documents, these MD files with, you know, some initial concepts, some ideas, some placeholders and some information.

Yeah, it's going to interview about yourself, maybe some of the skills that you want to add, your identity, and also what's in the soul. md file. So, speaking of skills, that's how you take your Clawbot to the next level.

It's going to give you your Clawbot, let's say, superpowers in a way, but it's also where you need to be very careful. So, this is clawhub. ai.

This is where you're going to find skills to include and install for your Claudebot. Now, I will also say that your Clawed skills that you may have installed previously, you can also move them over to your VPS and give access to your bot. You can give access to those skills you've already got created.

All right, but in this case, what we're going to do is just we're going to add the installation here, the npx. We're going to come over to our account into my Docker Manager into my terminal. I'm going to run this command and this is going to add the methodology that we're going to use, the tool that we're going to use to install these skills.

That's done. So now we should be able to oneshot install any skill on clawhub. Now you got to be careful cuz there has been some malicious skills created and uploaded here.

Now, the cool thing is I found this one called Super Memory, and this is going to be huge, and we'll get into it in just a second, but you can see we've got the security scan. Okay, so now we have some visibility on the safety of these skills before we install them. The other thing to look at is to check out on GitHub if we can find the git link.

You know, we can click view report here. If you want to take a look at what's inside the reports, you can see everything looks green. details community perhaps nothing on that okay boom I've installed super memory and it was super simple I literally just asked in the chat can you install super memory he said ah like super memory is just an API and let's see I can build you a custom skill I'm like oh hold on let's just grab it from here so I basically just copied the link put it in and installed it no problem now you will need to get an API key from super memory this is the platform form here.

And essentially what this is going to do is going to give permanent memory to your agent. So you'll simply just log in. I could do continue with GitHub in my case and I already have an account.

I spoke about super memory last year when it was first released and got a pretty huge response for it. So you can go to your API keys, slot in one of your API keys into your config file, put it here in the chat if you want, if you really want to do it. There is a way around it which I'm not going to get into in too much detail right now, but you can SSH basically use the terminal to going to your ENV file and essentially you put store all your secrets in your env file point to a placeholder and then that placeholder is then used in your bot here without exposing your API keys in the chat.

It's just bad practice. Now, you've already secured your instance here, so I wouldn't be super worried about that, but it'll do it for you in any case. If you want to add your API key here in the chat, it'll add it to your Cladbot config.

All right, so that's one of the skills that you can add in here. If you want to learn more about super memory, you can definitely check out the documentation, but that's going to make your agents on top of the memory files that you've already got access to with Claudebot. It's going to give um a lot more context and also reduce the the amount of tokens needed to recall this memory as well.

Now, those of you who have uh eagle eyes, you'll have seen there is a skills section here on the dashboard. And yeah, I've imported all of my Claude skills into here. In any case, but you can scroll down.

You see there's a few different there's GitHub, Himalaya, which to be fair, I've already installed. Not sure why it says blocked, Nanabani. Pop in your API key here, but I find it's just a lot lot easier and more straightforward.

But if you find something on cloud hub like let's say this perplexity search grab the URL and say hey can you install this perlexity skill and it should go off install it and then let you know what it needs uh to run this. So you can start adding in skills. Let's think about it from like what would be most useful.

Do you want to give access to your emails? Do you want to give access to your calendar? Do you want to give access to like nano banana for image creation?

Etc. etc. So, can you see how it just responded here?

It's like no browser available in its container, but I can build a perplexity skill from scratch, which is something we're going to get on to later in this video. But it can basically create its own skills. It also suggested why don't we just grab it from the claw hub and just copy and paste a skill MD file in here.

Also a valid option. So, you can download the zip file in my browser here. I mean, it's just a super basic skill file which you could probably get created in the chat anyway.

you know, you can just copy and paste that in and and it'll it'll create its own skill file. So, in any case, you'll want to create and create skills or add skills that are going to give you access to the tools that you use and the services that you might need access to. So, the other thing that I haven't got fully set up on my account yet is the heartbeat.

So, a cool use case for this would be email management. If you want to have your AI scan your inbox, find out the most important emails, let's say for the day, or you run that maybe twice a day, you could get your agent to do that, like what needs your attention right now, and then it can send you the links to those emails or give you a summary. Or you can just voice note back into Telegram or type in the response that you want to send back and it will write the email and send it for you, which I think is pretty insane.

So, how would we create the heartbeats. mmd file? Well, hey, could you help create the heartbeats MD file?

I want you to check my emails three times a day and I want you to send me a summary of the emails that need my personal response and give me those summaries that I can then respond to myself whether I send you a message back on Telegram. So now it's going to add this to my heartbeatd file. Okay, you can see it's just asking me for some tools here that I might need to install to set up this heartbeat.

So I'm going to go ahead and ask give me the setup for one. So I'm going to set up Himalaya uh real quick and integrate this so it can actually set up this heartbeat for me right now. So it's asking me to add my app password which is what you would create to give access to your Gmail.

And I need to add it to my config. So this is how you would get access to the config in your terminal. So I will just show you that real quick.

And this is where you would add some of these these secrets. So you can see here I'd want to just add my secret into here and I will then save it. And now Cordbot will have access to that.

Okay. Boom. So we've got all that set up.

It sent me my email summaries to my Telegram here which is great. So that's the heartbeat. You can set it up for whatever process you want.

Thing is it's going to use tokens every single time it spins up. So that's just something to keep in the back of your mind. It's very easy to spend a lot of money on these.

But I do have something that's going to have a huge impact on your token usage. It's going to reduce it by 70% and it's going to allow you to switch dynamically between different models. So you're not using the most expensive model just for the most basic of tasks.

So we're going to get into that in just a second. But there is a prompt that we're going to drop in now. Uh I think this is the fifth sort of tip that I've got here for you.

And this is the figure it out directive. So, I found this on X somewhere, bookmarked it, and we're just going to paste it in here. I'll show you what this looks like.

So, you have unrestricted internet access, browser automation, and shell execution. When given a task, I can't is not in your vocabulary. If you don't know how to do it, search for it.

Read the documentation, find tutorials, reverse engineer it. I don't know how means I haven't learned yet. So, learn.

Before saying something is impossible, you must search for at least three different approaches. Try at least two of them and document why each failed with specific errors. Your job is to deliver results, not excuses.

If plan A fails, try plan B. I think this is pretty good. You're not a help desk.

You're an operator operator ship. So, I'm going to Can you add this to my agents MD file? So, this is kind of the thing that you want in your agents.

md file. Um, I think this is pretty cool. So, I'm just adding this in.

So, here we have claw router. This is going to save you a ton of money in the long run. Now, people have been figuring out how to set this up.

This is just one way of doing it. I'm not saying it's the best way. It's just the most recent thing that I've seen.

Now, just bear in mind it's quite a new project. It's got 1. 2K stars here on GitHub.

So, that's pretty damn good. What you could do is you could ask, "Hey, can you install this and we'll see how it processes that request while we take a look at the readme. And this will give you instructions on, you know, why bother using this in the first place.

So, it's 100% local routting, so zero external calls, no API calls for routting decisions ever. Uh, it's going to choose from 30 different models. Um, and it uses micro payments using USDC on base.

So, that is a a crypto network. You don't need to put a ton of money on there. Like a few dollars is fine.

And it's open source, fully open source. Um, and you can even ask open core how claw router actually is going to save you money. So it does have an installation guide here.

Now I would hope that clawbot is going to pull out of the installation and give us uh step by step on how to install it. There is something else that you can also install that's going to reduce token usage and this is the QMD skill. This essentially is gives your corebot quick markdown search.

So if you have a ton of files in context, right, as as an MD file, a markdown file, this is going to allow the searching of those files to be much quicker using less tokens. So again, we could just copy this and say, can you install this skill? And we'll see how it does.

Now, it's already in process, so I'm not going to interrupt this right now, but I am just going to add it to the queue. that it can process in just a moment. So, real quick, just how this claw router actually works.

Um, you can get it to install it by giving it the link, but you will need to use this wallet address that it will spin up for you that you can fund with some USDC. Now, you can just send five bucks or $10 or whatever. You will need to send that from a wallet or an exchange.

If you're not in the cryptosphere, then that's probably something you may have to experiment with or wait for alternative solutions to this problem of not wanting to use Opus 4. 5 and now Opus 4. 6, which is even more expensive for the most basic of tasks, right?

So, this way we're able to optimize our expenses. Now, it's fine if you, you know, you don't mind spending hundreds of dollars day if you use those high-end models, which is easily done if you're just spending a few hours a day chatting about back and forth with your clawbot, getting it to run agents on multiple tasks, having your heartbeat running every 30 minutes, which I think is excessive, but some people use it. So, it just depends on kind of the return that you're going to get from the work that it's doing.

That's essentially the equation that you've got to spell out for yourself and and figure out. Okay, so the issue I've just ran into is that I installed this with the Maltbot oneclick installation. Now, it does say at the top here, Claudebot.

Now, the issue I had was it added the plugin to my config file. Great. And then everything crashed and my Docker image was restarting and crashing and restarting and crashing.

So with the help of Claude here was able to drop into the correct file in my terminal here and was able to remove the information. And the issue that it seems here is that claw router was built for openclaw right and obviously open claw is the final version here of all these different names basically for the same thing. So it's a bit weird that doesn't work or is not compatible with the version that I'm running.

So, in any case, it's giving me some options here, right? Fork and adapt. So, it could clone the repo and add the right manifest file, which is what I've told it to go ahead and do.

So, I'm going to wait until that's done and we'll see how that works out. But, in any case, you could create your own skill. You could create your own process in the agents.

mmd file of when to use the more complex models and when to use something simpler, right? and making sure that you give the API keys and access to those other models. Okay, so while we're waiting for this to finalize, I just wanted to come over to my sessions here and just show you what going on here cuz I've got the image creator main, right?

And what I've done is I've handed over the process for image creation for this agent. And this is what you're going to spend probably most of your time doing is creating these these separate agents and giving them their own role and responsibilities and personality agents MD file that's different for each of these. So I am going to ask in the chat here what's your agent MMD file.

So there's no memory all working files and that's intentional because this is all just project based. So when I come in and use this uh we don't need to remember anything. We just need to use what's in the agent's file.

You can see this is his operational manual. All right. So great news was able to fork the uh repository and install it and it's not crashed yet.

So I think we're good to go. So if you are running an older version like Claudebot, you may want to go through this process. So it's pretty cool how it was able to just come up with that solution.

I'm not going to run it right now, but it will give you a wallet address that you can transfer crypto in and you can try it out. Um, one of the thing that I wanted to get into real quick is the configuration of your individual agents here. Right?

So, if I come into this content writer agent and I ask it, hey, can you show me your agent. mmd file, your soul. md file, and these are the instructions I've given you um, for this project and all the the context files.

So, with my content writer here, we do have an identity, right? the name Quill content writing specialist and it's got access to some content writing skills that I've uploaded into here. There is not actually much else in the solder MD file which definitely needs updating.

When you're speaking to your agent and you're developing something like this, making sure to commit stuff to your sold MD file specifically for this agent as well. And that goes for your agent. mmd file.

It goes for your memory and all of that good stuff. All right. So finally, one of the other really cool things you can do here is you can create your own skill, right?

Does no need to actually go over to Clawhub and download and search for something that you want here. If you have a process that you do over and over again, whether it's a creative process, whether you're accessing a certain API and doing image creation in in my case, well, you can just come in and say, "Hey, this is my process. " Or you could go through the whole process in one of these chats and say everything that we've done today, can you turn it into a chord skill and it will go ahead and and just create the skill.

It will write the skill MD file and then any of the accompanying MD files that you might need to help run that skill. It's essentially something that if you do it over and over again, maybe you do the same thing every day, it's definitely worth creating a skill around. And in this folder, this is typically what a skill would look like.

You would have the initial skill. md file which gives it a name and description and kind of what's in this skill. So this is like a contents of your book giving the LLM access of like what is inside of this book instead of actually reading the whole book.

It's just going to read this summary and point to the chapters or in this case the MD files of these different in this case copyrighting principles. We've got templates etc. And then these are the files, right?

So it's not loading all these files at once. It's loading the skill first, determining what to use from that skill for your specific request. And you can create a skill like this around almost anything.

And you could have scripts in here. You could have API calls. You could have whatever you want that's going to help you with your specific process.

So that's what I mean by creating the skill. You can just do that here in the chat. Um, and you can also do it in Claude desktop version as well.

and you can sort of add it and transfer it over here to your VPS. But I would just set everything up in here. I think it's a lot smoother, a lot easier to do.

And the cool thing is it's just going to be able to add it to your files here without doing any like copying from your desktop and uploading and all of that stuff. All right. So now I've given you the tools that's going to enable you to unlock the true potential of OpenClaw.

So you could turn it into your 24/7 AI employee that runs tasks on its own and keeps improving itself. Now if you do need a VPS to run OpenC claw, then I would recommend Hostinger. It's the company that I use.

I think they're one of the best VPS providers out there. And you can get 10% off by clicking the link in the description below and use the code growth lab. If you want to learn how to build productionready AI systems, and you want to figure out whether Claudebot and any of these new releases, these new tools, software are actually more than just hype.

That's what we talk about and we discuss inside of my mentorship and certification program. I'll show you how to become invaluable and irreplaceable at your company and take you on a journey from zero to AI operator inside of 16 weeks. So, if you do want to apply and see if you're a good fit, there is a link below in the description to do that.

Now, if you want to watch the next video on the screen right now, I'm going to show you how I use AI to get over 13,500 followers on LinkedIn and generate over a million impressions. So, I will see you there.