Claude DESTROYED OpenClaw…

Key Takeaways

Claude have literally announced you cannot use Claude or off with open claw. And this has started straight away. So you can see here Boris from Claude Claude, he has said starting at 12:00 p.m. PT Claude subscriptions will no longer cover usage on third-party tools like open claw. And you can still use these tools with your Claude API, but that means extra usage costs, right? So I'm going to walk you through exactly what you can do and how you can get around this and and some of the best ways to use open claw without having to worry about stuff like this because there was a lot of people using Claude with open claw previously. Obviously it was amazing with the limits and everything like that. Obviously it couldn't last forever and I think it was mainly around costs and that sort of thing. So I'm going to walk you through some of the best free ways to use open claw instead of using Claude directly. Um we'll talk you through it. I'm going to start off with the first method which is using open router with Qwen 3.6 plus, right? Now this has a million token context window just like Claude. It's designed for agentic models and it's one of the most commonly used APIs on open claw. And in some ways this is better than using Claude because it is free. For example, if you were using Claude directly with open claw previously the problem with that was that you were still paying monthly for their subscription for Claude subscription. Now you can use Qwen 3.6 plus free as an API using open router directly inside open claw, right? So you can actually get access to this directly. Now if you're wondering how do I set this up, how do we get this to work, let me talk you through it now and I'll walk you through all these different methods and the different ways that you can use this to use these APIs for free. So one of the easiest ways is if you're if you've already got Claude code set up on terminal which you probably have because if you're using Claude and you're watching this video then you probably were you already have access to Claude code, right? So run Claude code inside your terminal like you can see right here. And then you're going to copy the documentation from Qwen 3.6 plus on open router. If you can't find it, you just type in Qwen 3.6 plus here, right? And then you go here and you get the information, right? Now what you're going to do from this section is you're just going to grab the information from the page, right? So select all. And then you're going to go to terminal and inside terminal you say help me set up Qwen 3.6 plus free see model details below with open claw installed locally. And then you paste that information here. Right? So you paste the information from the page which we've pasted in there and then we can allow Claude code to set this up for us directly. If you don't want to do that then you can actually go through the setup instructions inside open claw or another option is you can actually go to open claw directly and just say help me install Qwen 3.6 plus. Now you could use this just for sub agents or you could use this directly as a main agent. It is good enough to be the brain of your operation. It's a it's an API by Alibaba. It's free. It's got a great token context window. It's actually newer than pretty much everything available out there right now cuz it was just released on April the 2nd and you can get access to it, right? Um it was one of the first I think it was the first ever to do a trillion tokens in one day on open router. So it's pretty crazy. So that's how you can use this with uh open claw. Now if we just another little tip for you here get the GitHub from open claw just so Claude code knows what you're talking about. It should have persistent memory, but it often doesn't, all right? So I'm just going to say use this to help you set up open router with this, right? But that's how you can do it. So that is method number one. I just want to recap on this because let's make it easy for everyone. You have many different options. And option number one is Qwen 3.6 plus on open router which is free. So that's how you can do it and you can set that up in Claude code or open claw directly. Or you might even be able to use it inside the onboarding process as well with open claw, but I've not tested that out and I haven't seen anything from open claw saying they've updated that. So those are the two options that I would focus on directly, right? Now let's move on to the next option. And let's see what we got here inside the comments first of all. Let's see what we got here. Can't wait for this one. Thanks Brian. GLM 5 I would agree. Yeah, GLM 5 is another option that I'll come on to in a second. Uh 6 months from now it's going to be insane. It's it's getting crazy by the day. Uh Brian says I've been using Minimax 2.7 high speed, but I'm not married to it. If Julian says it's better, I'll switch tonight. Yeah, I would definitely just try everyone has their own preference on APIs, but I would try Qwen 3.6 plus if you want as well. All right. Uh Skido says I'm trying to get paper clip to work. Now if we can just invoke VS code extensions with our tools, we can keep better track of all the sessions working together. Nice. And then Woolsey says cheers Julian, happy to help. Family says bro stop hating on open claw. I'm actually trying to help people using open claw. Not sure what that's about. Uh if we already have Claude code subscription, can I also switch to open router as well? Yeah, that's that's exactly what you do, right? So just like I was showing a second ago, you get Claude code to set up open router for you and then that's how easy it is, right? Um and then it can it can set up Qwen 3.6 plus as an extra list inside your models, right? And so that's that's an easy way to do it. Now I'll I'll show you an example what this looks like in practice because Claude code has actually finished setting it up. And we can change it to default models as well if we want to. Tom says love your content G. Thank you so much, sir. All right. So if we go to terminal here we should have this running as you can see. So we can get Claude code to set it up if you want, but you can see it's added inside the config, right? So if we it let's just test this out and just make sure that Qwen 3.6 plus is actually working with open claw. So we'll just run this now. And then if we have a look through the list, we've got this working here. Yeah, so you can see we've got Qwen 3.6 plus free working inside open router. So that's how easy it is. Let me just check it's working. Boom shakalaka. Look at that. We got it working, my friend. So Qwen 3.6 plus free with open claw, that is what you can use now, right? So if you if you were like oh no Claude blah blah blah, no no don't worry. Don't worry, my friend. You got Qwen 3.6 plus free and that is probably a better API, right? Because it's free. It's got a million token and it's newer than the Claude, right? Now is it actually in reality going to be as good as Claude? Probably not, but is it designed for agentic systems? Will it work as good? Yeah, for sure. For sure. You can see how quick it was to respond, right? So that is method number one. Now I want to show you a couple of other methods. If you if you if you're insistent on testing something else out or if you want to run sub agents as well, so another option that you have, right? Is you can go inside Ollama and Ollama has cloud models. Now if you were using for example Claude with open claw, you were paying minimum like $20 a month, right? To use this. Now what you can actually do is you can switch over to Ollama cloud instead like this, right? So for example, it's really simple and easy to do this. So you just number one you sign up to the Ollama cloud plan. The reason you want to sign up to the plan and not use the free model is like if you're going to use this as your main model inside Ollama then you're going to hit limits eventually, right? And so to avoid hitting limits on your cloud usage, just sign up to the pro. It's like $20 a month. So it's the same as Claude and you also get a choice so you can switch between Kimiko 2.5 cloud. You could use GLM 5 cloud. You could use Minimax M 2.7 cloud. You could switch between them to see which one you want to prefer to use. But, the main thing here is like you don't have to rely on Claude to run your Open Claude, right? You can switch between these as well. And it's the same price as Claude. So, let me show you how this would work. So, to make the switch, you just go to your terminal and you would type Ollama launch Open Claude like this, right? So, if we open this up, I'll open up a new tab here. Ollama launch Open Claude, and then we switch between the models that we want to use. So, for example, if I want to use MiniMax M2.7 Cloud, I can. Now, bear in mind, when we do this, we just open you just press up and down to switch between the models, right? So, the recommended ones are Claude 3.5 Cloud, Kimiko 2.5 Cloud, MiniMax M2.7 Cloud, and there's a bunch of others, too, but I would go with Kimiko 2.5 Cloud, for example, right? So, if we switch that now, that's going to launch um that's going to launch directly using Open Claude now, right? For us, directly. Now, another way you can do this is if you go to Ollama like so, and let's say, for example, you want to use Kimiko 2.5 Cloud, right? With Ollama. You can go to Kimiko 2.5 on the models list of Ollama. Make sure you've got Ollama downloaded, and then you run it here. So, you go inside your terminal and you make sure you've got Ollama Cloud set up like this. Then you can press {forward slash} buy to disconnect from it inside terminal, and then you're going to run this command to run it with Open Claude. So, you copy and paste that into your terminal, hit proceed, wait for it to load. Boom, right? We're in. We're in. So, we copy that now. And we're in, my friends. So, now we've got Kimiko 2.5 Cloud on Ollama, right? Now, if I say, "You working?" and we just check it's working. Boom shakalaka, right? That's another option. So, just to recap, you got Ollama. You got Option number one is Claude 3.6 Plus on Open Router. That's the best way. That's the best option by far, right? Then you got Ollama with the cloud plan. And that's like, you know, that's $20 per month. Let's talk about more options here. So, let's see what we got here. Oh, so it's shows on the list. Yeah, when you So, you don't even need to type anything. Like, if you go inside Open Claude, as long as you're updated to the latest version, you'll have the latest dashboard inside the the gateway, right? And so, you'll actually see there's a bunch of different agents I've set up. So, these are all different instances of Open Claude. And then on this drop down, we can switch between the APIs, right? So, I can choose between Kimiko or MiniMax or GLM 5 Turbo, whatever I want, right? That's how we switch between the models. Why should we use Open Claude when you have Claude Code scheduled runs, dispatch, and all these new features? Because they're just not that easy, right? Like, I don't know anyone using dispatch that much for Claude Code. It's just not that easy to use, right? Like, it's a lot of work. And if you use something like Open Claude, it's way easier and way faster to get it working on your mobile, right? Like, you can easily connect this to Telegram. You can't really do that with Claude Code. So, in many ways And also, bear in mind like you have to have Claude Code running on your desktop, which I don't know about you, but I close everything as soon as I as soon as I stop using it. So, like, for example, Open Claude is always running in the cloud. It's always ready. Same with Hermes. It's always ready whenever I need it. And I don't need to like have like this big program running in the background. Whereas, for example, if I want Claude Code to run on my mobile, it's number one, not going to be that convenient. Number two, it's proven only to work like 50% of the time. They explain that on the website directly. Number three, I have to have a big program running in the background. I have to have Claude Code desktop or Claude Code inside terminal running in the background, and I don't like to do that because it's just kind of like it's just a lot of work. Whereas, for example, with Hermes, I can shut everything down and it's still running in the cloud, right? It's still running 24/7. So, it's just much easier. It's a much more smoother integration. And then also, bear in mind like with Claude Code directly, it's not as free, right? If you if you're using an open source model, you can get a lot more technical with it. You can do a lot more interesting things. Whereas, with Claude Code, they are going to stop some stuff, right? That you can't do with Claude Code directly. Let's see what we got here. And will the custom models inside of Claude Code be able to summon sub-agents, agent teams, or can I only code it up? Yeah, like So, you don't need custom models inside Claude Code to create sub-agents. There's already a built-in feature for sub-agents. So, you just use the sub-agent feature. Vote says, "Happy Easter, champion." Thank you very much, sir. Happy Easter to you, too. And then Sergio says, "Claude Code is excellent for building apps, websites, APIs. Open Claude for agents." Sergio nailed it. That's exactly it. That's exactly it. Yeah, [snorts] nailed it. All right. So, next option. You can also use Atomic Chat. So, this allows you to download Atomic Chat, which means you can run models locally free forever. Or you can plug in your API from Claude 3.6 Plus, and then you can use that directly with Open Claude. So, that is method number three. The other option that you have is you can use GLM and the coding plan. So, the coding plan of GLM supports Open Claude, right? So, you can use Open Claude with GLM, which is a really good API as well. If you want to set it up, you just follow these instructions. If you want to get the coding plan, you just sign up. I think it is pretty cheap, if I remember correctly. Let's have a look. It might have the pricing here. I mean, let's have a look at the cost here. Yeah, as you can see, it's it's pretty cheap here, right? Like, if you switch to monthly, that's what you getting. And you get three extra usage of Claude Pro plan, right? So, that's a good option as well. I mean, for that price, wow. Wow. Yeah. So, that's another option. So, you got GLM coding plan. You got Atomic Chat. Atomic Chat is only really good if you're like if you've got a good setup, right? Otherwise, you have a need good local hardware to run the local models, or just plug in your Claude 3.6 Plus free API, right? And then the cost equals the cost equals free. Right? So, these are some good options now. You got Claude 3.6 Plus on Open Router. You got Ollama with the cloud plan. You got GLM with the coding plan. You got Atomic Chat. And then there's one final method, which is MiniMax. Let me show you how that works. So, this is MiniMax on the coding plan. You can connect it with Open Claude, and you can use this subscription to use that, right? MiniMax is a really good API as well. It's actually so good that 30% of it was built by the AI agent itself, right? It's self-improving. It can do its own thing. So, if you're wondering how much it costs, here you go. Here's the details, right? As you can see the pricing right here. Um that's that's your cost here you're looking at. So, the coding plan, you can see the usage limits and everything else, right? So, again, it starts at $10 per month. So, that's five solid options that are all very affordable or free for using Open Claude. You got Claude 3.6 Plus instead of Claude. That's probably your best option. You can use Ollama if you want to. You got GLM coding plan. You got Atomic Chat. And then you also got MiniMax coding plan as well, right? There's other options, too, but I mean, five options already is a lot, right? Like, you you don't need more than that. So, thanks so much for watching. If you want to get all of my best trainings on Open Claude, how to set it up, and amazing community of 2,700 business owners using AI agents to grow, save time, and scale, check out the AI Profit Boardroom. Link in the comments and the description, or go to the air profitable body.com. Inside the calendar, you can see that we have weekly coaching calls. Inside the classroom, you get all of my best trainings. Inside the map, you can connect with local people in your area. And it's got everything you need to win with this stuff, right? Now, let's see what we got on the questions here. Brad says, "I'm too scared to use open claw if I try to add a llama." That's up to you. I mean, like you can It's up to you. Like if you feel that way, then probably just trust yourself. Um what else we got here? How do you generate images and videos free with open claw? Um we've actually got a training on that inside the air profitable body, right? Uh images, you can't generate for free, but videos you can't generate for free with open claw. We've got training inside the air profitable body on how to do that. Uh we've got another question here. Isn't it easier to just get an API key through open router and then tear what kind of model you want to use based on the task you use at hand? Uh I wouldn't say so. Like you can use auto open router. That's an option for APIs inside open claw. But if you've got a choice between using the free models, right? AKA Quen 3.6 plus, or you've got a choice between switching automatically, which could get very expensive if it switches to Claude, why would you do that, right? And so you want to go with the free models because the free models are free, right? And if you want to do this in in the most cost-efficient way, which is what people are doing with Claude previously, then you do it that way instead. Have more deeper thinking tasks be done by Claude and then the easier ones will be done by Quen or something. The thing is if you use Claude as API and you're paying for that, it's going to get crazy expensive, right? And most people don't want to do that. So I'll give you an example like, yeah, no, it it gets expensive. Um let's see what else we got here. Aidan says, "Been using open claw for 3 months. It's great, but sometimes memory hang and just get stuck. Have you encountered that? Can we fail back to another open claw instance that has different API provider option?" Yeah, you can. So like inside open claw, um you can you can just switch between APIs, right? So if you want to have one as a failback, just switch between APIs like this and just have multiple ones. That's what I do. Do you have instructions on how to set up Telegram with topics for each agent? I'm talking to all my different agents in one thread in Telegram. So you would just start a group chat and then you type in {forward slash} topics. That's It's as simple as that. There's not even like a You can do the same thing inside Discord as well. I've got a Discord training inside the air profitable body on exactly how to do that, but yeah, the main thing is you can You can have I I actually prefer the Discord method because then you can have separate threads and that's much easier to organize with your open claw than it is with Telegram and topics. How would you build a full marketing team with open claw? Yeah. So if you have a look here, you can check out our training inside the air profitable body, link in the comments and description, to check out our training on how to set up a marketing team with open claw, right? We've got a full guide on it right here. So just check out that video if you want to learn how to do that. Open claw gets stuck sometimes when where you hitting 200 sessions, for example. Can we fall back to another open claw instance to heal the active open claw that is stuck? So in those situations, what I would do is just use Claude code to fix me, right? So for example, if I have problems where open claw is not responding, instead of like falling back to another instance, and I don't think that works for I've seen I cuz for example, I've got like multiple sessions with open claw. As far as I'm aware, it's not going to fall back to another uh instance. So we've got all these different instances here, but it won't fall back if one of them fails, right? It can fall back on the API, but it won't fall back on the instance of the agent. Although if you do ever struggle or you need to fix open claw with your local setup, then just go inside Claude code and ask it to fix it for you. Will open claw or paperclip know how to use a better thinking model for planning or review and another less expensive model like Kimiko 75 to implement tasks? Yeah, so you can train it to do that. So you would tell it, "Okay, use this model for thinking and this model for sub agents." And you just train it to use like that. The other option that you have is inside open claw, you can use open router auto. And with auto open router, it will automatically switch between the different models depending on the complexity of the task. So that's another option. I have to remind open claw to remember yesterday's memory. Why can't it remember? Cuz the the memory is not very good. Like this There's plugins you can actually get for this, right? So if you actually go to There's so many different options for memory inside open claw if you want to improve it, right? So you've got your memory MD file as well. Sometimes a lot of people are using Honcho. I've seen this for Hermes as well, Honcho memory. And that's a good alternative if you want to improve the memory of your AI agents, right? So you can integrate this with all your agents. It doesn't just have to be open claw and then the agents will be better. The The other thing that you can do is you can do like {forward slash} new or {forward slash} restart inside open claw. And what that would do is refresh the context of the session because if you're running all your tasks inside one chat instead of separate threads, then it's too much context for open claw to be able to deal with. And so what happens is it forgets stuff that it's done previously. It forgets feedback that you've given it. And so if you want to have like a better setup for that, just have separate threads, restart the session, compact the session, start new sessions as well, and then use something like Honcho memory inside your AI agents and you'll get better results. So that's basically it for me. Thanks for watching. Check out the air profitable body if you want to get more training on stuff like this. Today, I'm going to show you how to use Atomic chat to run open claw free forever. Now, this is an app you can download and it sets up open claw in one single click. It is the easiest way that I know to set up open claw. And I'm going to show you exactly how to do it today. So if you're non-technical or if you want a free setup for open claw or if you're just curious about learning new stuff about new agents, check this out. So what you want to do is download the app from this website like you can see. Then once you've downloaded and installed it, you're then going to open the app like so. Once you've done that, it's going to load like this and you can see here that we have dashboard, AI models, and skills, right? So what you want to do is go to the AI model section inside Atomic chat and then go to local models. And inside here, you can switch between local models that you want to use. So for example, you could download GLM 4.7 flash and run that with open claw. You could use Nema tron 3 nano. You could use Gemma 4B. We've actually got Gemma 4B running right here. So if we click on activate because I've already downloaded that, it's 17 GB. And if you don't have a powerful setup, don't worry. I'll show you a lightweight way to set this up in a second. But now we've got this running, we can now go over to dashboard inside open claw, right? So you can see we've got open claw running inside the dashboard here. And so what we want to do is we go over to the chat. And now we have our AI agent running with Gemma 4 inside open claw. That's how easy it is to set up. So if we say, "Hey," and this will respond with Gemma 4 using our local model. And this was set up in one single click. So number one, we can run open claw free forever with Atomic chat because we're running local models. Number two, we can get it all set up in one click. And number three, it's super easy to set up, right? Now we've got all the other stuff running inside here. So you can see it is free. The costs are added up here and it's free. Um and we can see like the event log and all that sort of thing. We can see the schedule, the instances, the channels, the skills, the agents, the nodes. Everything that you get inside open claw, you get inside Atomic chat. And you can see it's responding to me as well here, which is great. And it responds within a minute, right? So we sent this message at 9:59 and it replied at 9:59. It is a bit slower than using your normal models because it's running locally, but that's basically how you could do it, right? Now, it's really easy and simple because all you've got is this. And then you pick your model and then you just click active, right? So we could download any of these and then activate them whenever we want. Now, also another option that you have is you can get an API key from OpenRouter, paste it in here, and then select the model that you want to use, right? Now, you can see all the different models inside this chat, right? Some of them are free, some of them are paid, but you bring your own API key, so you can choose which model you want. Now, if we actually have a look here, I can't actually see um Qwen 3.6 Plus. Let's have a look here. We've got Qwen 3.5 Plus. I can't see Qwen 3.6 Plus inside the list. I just want to check something here. Qwen 3.6 Plus is is, by the way, the new update. So, let's see if we can update it here. So, you can see all your skills inside your settings as well. So, like, for example, self-improving agent. That's something You've seen how many stars the skills have and how many downloads they have, too. You can check out all of the models that you have, and you can connect your messengers to it, right? So, you can connect Telegram to your OpenClaw pretty easily using this method, right? Um you've also got voice recognition, which is pretty cool. So, you can speak to your chat directly. And then you got some other options here, too, right? Um and you can create backups as well. So, you can choose your OpenClaw folder, you can choose your agent workspace, you can choose whether you show it in the sidebar or not or in the terminal. So, if we click on that, you'll see that the terminal opens up in the bottom left, right? So, we can toggle that on and off. We can switch between command approval, um whether we want to send statistics as well, right? I like the create a backup section. I think that's great, cuz a lot of people worry about losing their um you know, the back the all the all the time they spent. Right now, it's only inside Telegram, but I think that's good enough. You can connect inside Slack as well, but I think Discord, WhatsApp, Signal, iMessage, etc., they're all coming soon to AtomicChat. And then again, you can switch between API keys here. So, you could, for example, like you could grab an API key from Ollama, and then just run local models, right? Uh so, you can see here, for example, we don't have to use the local models inside inside um AtomicChat. We can switch between the local models inside here as well, right? So, if I wanted to use this free again, another option that I have is I can just switch to Kimiko 2.5 Cloud as a local model, right? And then if we go to a new task, and we go to dashboard here, then we go to the chat, you can see this is now running with Kimiko 2.5 Cloud. Now, we can switch between any of these APIs if you want to, but if you want to run this free, you just do it like so, right? And then we just say, "Hey." Um and that's basically how we can do this, right? And so, it's quite easy to set up. Now, if you want to run cloud and local, this is how you do it. And you just need to grab an API key, right? Now, you can create an API key from Ollama to connect it by creating that inside your dashboard here, right? So, let me show you how that works. So, we go to our Ollama Ollama dashboard. Settings, keys, and then we just grab an API key here. So, let me do that. So, if we go back to Atomic now, I've put in the API key, we hit save here, we test the connection, and you can see it's connected to Ollama, which is great. So, now we can use the cloud models from Ollama, right? So, if we go to dashboard here, then we go to chat, and we say, "Hey." Boom shakalaka, right? It's working, right? Um by the way, when we didn't have the API connected, it didn't work, so you can see here it fails. When we do have the API connected, it works and it's using Kimiko Kimiko 2.5 Cloud. Now, if you do not use OpenClaw much, this is free as long as you stick within the token limits of Ollama, right? So, you got a couple of options there. Or you got quite you got quite a few options in terms of using this free forever. So, you've got the local AI models with AtomicChat, right? You can check these out here and just use whichever one you prefer. You can go to your API keys, and you can add an API key for Ollama, and just stick within the token limits, and then switch to Kimiko 2.5 Cloud or Minimax M2.5 Cloud. Or, if you hit the limits and use this a lot, you can switch to local models inside Ollama, and you just switch between the local models here, right? You download the one that works for you best for you. So, these are all like simple ways you can use OpenClaw free forever. Um and it's it's easy to set up. I I do like AtomicChat as well, because it's just so easy to to get OpenClaw installed, right? It's just a one-click setup. It's free to use, it's free to download. You have a nice little designed app right here. It's easy to navigate as well. Um so, that's the way that I would do it, right? And that's the simple setup guide. Now, if you want to get more training on how to use OpenClaw, how to set this up, how to run local models with OpenClaw, check out the AI Profit Bot in link in the comments description, or just go to the AI Profit Bot in dot com. We've got an amazing community of 2,700 business owners who are all learning, using, and deep into AI agents like OpenClaw and Hermes. You can go inside the calendar, get weekly coaching calls, you can go inside the classroom, get all my best trainings. But the main point here is this is just an amazing community to learn this stuff. Plus, you can personally connect with me. So, I answer the DMs personally. You can connect with me, you can ask me questions. Um I can get to know you, and you can ask me questions whenever you need help and support, all right? So, that's the the best place to go. Let's see what questions we've got over here. So, I don't says, "Tip, start a project and tell OpenClaw to sync everything related to that project in that folder." Nice. Thanks for the tip. You can run OpenClaw You can run ClawCode free now. Yeah, you can do that with Ollama. I've got a quite a lot of tutorials on YouTube about how to set up. Brad says, "Atomic is easy to set up." I would agree. Brian says, "Where was this 4 weeks ago? Would have saved me 4 weeks of 12 to 14 hour days." Yeah, I would agree. I would agree. It's getting easier and easier to set this up. Use OpenRouter. Yeah, I I saw, like, for example, Qwen 3.6 Plus is now commonly used most commonly inside ClawCode and also OpenRouter, too. How do you add custom skills to AtomicChat? So, you just go inside the skills section, right? Um So, you go to uh add You go to skills, ClawHub skills, and then add custom skill, and then you can upload your own skills from there. So, that's how you can set up. Today, we're going to be looking at how to use Hermes with a custom memory. So, there was a new update yesterday from Nous Research that says Hermes 0.7 is out now, and the main headline here is that memory is now an extensible plugin system. So, you can swap in any backend or build your own memory inside Hermes agent. Now, why is that useful? Well, basically, what it means is that you can have this shared memory across all your AI agents, you can have a better memory, you can give better context. And the better your agent's memory is, the easier it's going to be used, right? The easier it's going to be to set it up as well. Now, if you want to set this up, you can use, number one, just update to the latest version, and then use the terminal command Hermes memory setup, right? Hermes memory setup. And what it actually supports is a bunch of different updates and memories like you can see here. So, for example, you've got Honcho. And these are all These are all like different ways of using your memory. So, you've got Honcho memory, Open Viking mem, hindsight, holographic, retain, and Byte Rover, right? Byte Rover is massive massive on GitHub. Uh people are absolutely loving it. Uh the one that I want to try out today is Honcho, but there's many different ones you can do, and I just want to see what each one does, and, you know, we can explore this together. So, it's the first time I've set this up, and we can just learn how it works together, and uh get this set up from here, right? So, this is a pluggable memory provider interface. What that means, essentially, is like you can You can take that memory, you've got it stored, and then you can run it anywhere, right? And so, it it stores and remembers stuff. It's kind of like, for example, uh having a USB memory, but it's for your AI agents, and you can just plug this in directly here. All right, so we're going to learn how to use this and how to set this up. Let's see what we got on the questions here. Nile says, "Amazing. Happy dope." And perfect says, "Are you AI avatar?" No, I am not today. I don't think um I did a video about it yesterday, but there's only one option I've seen for that, and it doesn't it doesn't work with lives, right? So, it's um the real Julian Goldie speaking right now, my friend. All right. So, let's get straight into this, and I'll show you exactly how to set this up. So, we're going to go into Hermes, and then we're going to start working on Honcho memory for AI, right? Now, if you've never used this before, good, because I've never used it before, right? We're just going to learn together. I want to learn this, and that's why I'm doing a video about it, right? And so, I'm not like the master on Honcho memory, but I do want to learn how to set it up how to get it working, right? So, I think there's two ways you've got this, right? You got basically you have the memory on the website, or you can run the GitHub as well. So, Honcho is available on GitHub as well. It's got 1,600 stars. Uh so, it depends how you want to set this up. I'm going to do the cloud version instead. And so, if you go to launch app here, we can log in like so, and then I can just log in in the background. So, let's do that. And the good thing about this as well is like it's focused on continual learning. What that means is like it's self-improving. That's exactly what we want, right? We want it to get better all the time. So, the challenge is memory systems are not that good right now. The solution is this continuously learns, and every single message triggers a new way of storing a new memory, right? And then also the good thing about this is you can plug it into Open Claw, you can plug it into Hermes, and whatever comes next in the future, bear in mind like Hermes and Open Claw aren't the only AI agents out there, and there will be something new in the future. Whatever comes in the future, you've got a memory that you can plug into that and and give it context straight away, right? That saves a lot of time. And also, you could store this across your team, right? So, let's get started with this now. It says, "Describe your project." I'm going to say running the AI Profit Boardroom community. Then we'll go from here. What do you hope to get? Ah, I hate these setups. We'll call this Goldie Agency. Add team members as well, which is pretty cool. And then we'll set up the organization like so. Let's have a look what the I don't truly understand the difference between like the GitHub. I guess they've just made it open source, is that why? Or you you store files offline as well instead of online if you don't want to run it by the cloud. And then let's have a look how this works. So, it does look like it's kind of you have to add a payment method here. Um but let's see how it works. So, we can plug in an API key. And then we can explore stuff. And we can go from here. Let's try let's learn how to use this. So, we're going to click on open docs here. You got four different layers here. You got the workspace top level containers. You got peers, any entity that persists but changes over time, so your agents. Your sessions, which you have threads. And then also messages, right? And that's how it works. And then I think you get Yes, you get this amount when you sign up, which is good. Perfect says, "What is the real use case for Hermes? Is it just vanity to see how AI agents work?" No, the real use case is like automating whatever you need to, right? So, like for example, for me, like I use Hermes to automate my social media, to create a lot of marketing content, to analyze my competitors. I've done a lot of videos and tutorials on that, so like you you the way that you look at use cases, a lot of people ask, "What's a use case? What's a use case?" The way that you look at use cases when it comes to AI automation is you look at what do you spend your time on? That is your use case, right? So, whatever you spend your time on, that's where you start. Now, what I spend my time on is very different from what you spend your time on. Everyone is different. And so, you want to look at, okay, where am I spending my time, and then how can I automate that using these AI agents? And so, that's the method that you want to look at. Don't don't just look at like, "Oh, okay, what is everyone else doing?" Look at, okay, what do I do day to day, and then how can I set that up directly, right? Now, if we could get the models 4 to 5x faster, so you can use turbo models, right? So, for example, GLM-5 Turbo is a way to run faster models with AI, so that might be a good option for you. And then Mayve says, "Are you switching from Open Claw to Hermes?" I'm using both right now. I use both cuz I think it's important for me as someone who educates people on how to use both, I need to be using both to learn day to day. Um and I would recommend like if you have time, use both. If you don't have time, pick one. If you're not technical, don't use either of them, use Claude instead, right? That's the way that I look at it. That's the advice I give. So, we've now got the instance set up for Honcho like so. So, you've got that set up now. Let's have a look in the explore section, see what we got here. So, now we can create a workspace ID, and then we can go from here. Now, if we go inside the API key section, we can create a new API key, and then we just give that to Hermes or to whatever we want to use, right? So, I'm just going to do that in the background right now. Perfect says, "Can you please share the link? I'm a member of your community." This is for Honcho, so you just go to honcho .dev, right? Um if you're looking for Hermes training, you go inside the AI Profit Boardroom, and you search Hermes, right? So, if you go inside here, search Hermes. In fact, if you go inside the AI Profit Boardroom here, you go to classroom, um and then go to Where is it? This section. We actually have a full 2-hour course on Hermes. So, I would start there, and I would learn. We've got a full step-by-step SOP as well on exactly how to set it up, what the use cases are, how it works, how to use it, and what you can do with it, etc. I would start there, and that's some really good training. So, just go to classroom, and then go to SOP updates April, and then Hermes, right? And if you want more training on Hermes inside the AI Profit Boardroom, type it in at the top, and you'll see all these different trainings, so you can choose which one you want and which one you want to learn from. All right, so we're just going to go back to this. We're going to grab an API key. Create one inside the settings section. And then I think the way that we would run this is we need to go inside our terminal. So, let's go inside our terminal here. And then if we go to Hermes memory setup inside terminal, we can get Honcho set up. So, let's try that. So, you can see that I'm setting this up inside the terminal here. All right? And then we choose which one we want to use, right? So, we can switch between all these different memories. We can use a built-in memory, the memory MD, and that's good, but it's not the best. The reason that they've set this up is because they want to give people more customization on the memory, right? And so, if we use Honcho, we can plug this in. I've just added that, right? So, leave that blank. And then, if we start a new session here, and we say, "What memory are you using?" Actually, I'm going to switch back to a different model. Yep, it's working. All right, and then we'll say, "Do you have Honcho memory set up?" All right. And then we can add it with information about us, right? So, we've connected Honcho, but it's empty because I just created a new account. So, I'm going to say, "Yes, set that up." And the good thing about this is we can switch between Honcho on Open Claw, and also on Hermes, and also on any other AI agent that we use in the future. And that's good because now we can sync the memories across the different agents. So, if we're using one or switching between them, no problem at all. And also, this is cloud-hosted, which means that, for example, if you're ever having a problem, and you have to fix Hermes, or you have to fix Open Claw, or you have to restore previous versions, well, it's not a problem because you can now switch and import the cloud memory that you have previously, right? So, it says, let's have a look inside the chat here. All right, it says, "Done. Populated it with the top five facts. You can now ask me things like, 'What do you know about me?' and I'll pull from Honcho memory." All right. So, let's go back to Honcho now, which we've got here. And if we go to explore, he's created this Hermes file, right? And so, he's added information about us. Now, obviously, like, that's not a lot of information, and not a great amount of information, but you see how we can quickly update this and we can sync it. And so, like, the the cool thing about this as well is like, we can add memory ourselves, or we can send that memory over to Hermes, or Hermes can add memory to Honcho, which means if we go out to Honcho and then we sync it with Open Claw, everything works together. Let me show you what that looks like in practice. All right, so, for example, you've got Honcho here, hosted in the cloud. This is your memory. And then that can sync to Hermes, and then you've got Open Claw as well. And these can sync both ways. Like so, right? And also, you can kind of sync Open Claw and Hermes together as well, because if Hermes updates Honcho, well, that updates Open Claw. And if Open Claw updates Honcho, that's going to update Hermes, right? Because they're all running from the same memory, as long as you've set that up. So, it's a really good way to just sync memory across all your AI agents, and it's also a way to get a better memory. Like, someone was complaining on the video before, like, your AI agents have to get stuff all the time. And so, this memory section, it makes itself learning, it makes itself improving, it makes it better. And that's what we're talking about here. That's how we can do it. So, thanks so much for watching. If you want to get more training on Hermes, how to use it, how to get the most of it, how to get the most out of Open Claw, get all my best trainings inside the AI Profit Bootcamp. And this is an amazing community with 2,700 builders who are all focused on growing their businesses and saving time with AI automation. It's a very active community, lots of cool people posting, lots of cool stuff, people sharing awesome stuff that they're building. On top of that, inside the calendar, you get weekly coaching calls where we go deep on this sort of stuff, and you can ask questions in real time. Inside the map, you can connect with people inside your local area or city if you want to meet people who are doing similar things to you. And you can also search for anything that you want. Plus, you get all my best trainings on this stuff. We update it every single day. We have a full 2-hour course on on Hermes. We have a full 6-hour course on Open Claw. We run new trainings daily showing you exactly how to use all this sort of stuff. And that's all inside the AI Profit Bootcamp. Link in the comments and description, or go to the aiprofitbootcamp.com to get access. Let's see what questions we got. I think we have we have no questions. All right. All right, we just got one from Sportis. Good to see you, Sportis, again. He says, "Sup, goat." But that's pretty much it. I don't think we got any questions on that. All right, either I lost everyone, or I explained it so well, nobody had any questions at all. Either way, let's move on to the next topic. Today, I'm going to show you how to use Gemma 4 with Hermes so that you can run Hermes free forever using Gemma Gemma 4 and local models in general. Now, Gemma 4 is Google's latest small, open-source, lightweight, efficient model, and you can now run it directly with Hermes, and this is free to run locally. Plus, it's very lightweight, and it's easy to set up. So, I'm going to show you exactly how to set this up with Gemma 4 so you never need to pay for an API again. All right? So, let me show you exactly how this works. Hermes agent, make sure you've already got it installed. If you don't know how to do that, follow the documentation here. Then you're going to have Gemma 4 running, and I'll show you exactly how to set that up now. All right, so, here's what you're going to do. Number one, make sure you have Ollama downloaded. If you want to download Ollama, you copy this command and run it inside your terminal. From here, go to models, then Gemma 4 on ollama.com, and you're going to run this terminal command here to set up and install Gemma 4 locally for free as well. So, just to recap, you downloaded Ollama. You have installed it via Gemma 4. And then, what you're going to do from here is you're going to run this command, all right? Hermes model. So, if we go inside our terminal here, we're going to start a new chat. We're going to run Hermes model like this. And then, from here, we need to set up a custom endpoint. All right? So, you can see this list of different models that we can use, right? We're going to go to custom endpoint, where you enter the URL manually. Now, make sure you have Ollama running in the background, make sure you've already installed Gemma 4. So, you select custom endpoint, right? Terminal Hermes model, select custom endpoint, then you're going to copy this URL like you can see, and paste it into terminal like this, right? Boom. Leave this blank, or don't leave it blank, and just type in Ollama, right? So, Ollama like that. Once you've seen that, it's going to give you the available models that you can run with Hermes. So, for example, let's say we want to run number two with Hermes, right? Minimax M2.7 Cloud. Let's say we want to run that. We type in two here. We leave this blank. And now we run Hermes like this. Boom shakalaka. We got Minimax M2.7 Cloud running for free with Ollama. Now, if we wanted to run Gemma 4 with this, here's how we do it. Run Hermes. Model. We go to custom endpoint. We take the URL. We add the API key as Ollama. And then you're going to select number one, Gemma 4 latest. We leave that blank. We run Hermes now. Boom, look at that. Gemma 4 latest running with Hermes. That is the easiest, cleanest setup I've seen for setting up Ollama with Hermes. I've not seen anyone else on the internet explain it um as simply and as easily as that. And that's how it runs, right? So, you do terminal Hermes model, select custom endpoint, use this URL. API key equals Ollama, and then you pick your model. And then just run Hermes, right? And that is basically how to run this directly with Ollama. And then you can run Hermes for free forever using Gemma 4. Or you could use one of the cloud models. The cloud models will be better, but there's limits on them. The local models are 100% free and no limits, but it depends on your setup. That's basically how you do it. So, thanks so much for watching. If you want to get more training on how to use Hermes with AI agents, how to use local models, how to get the most out of Hermes, all the best use cases, we have a 2-hour course on exactly how to use Hermes directly here to save time and grow. If you like AI agents, we have another full 6-hour course on how to use Open Claw here. We add new daily trainings as you can see inside the SOP section here. Right? So, for example, we show you how to set this up with Ollama and Open Claw. Yesterday, we ran through the new 0.7 update from Hermes. And this is inside the AI Profit Bottom community where you can ask questions. You can get help. You can get support whenever you need to. You can also go inside the classroom and get all of my best trainings on AI automation. You can join the calendar and jump on weekly coaching calls. And you can also check out the map and connect with people in your local area. So, that's all inside the AI Profit Bottom. Link in the comments description or go to the airprofitbottom.com to check it out. Thanks so much for watching. Let's see what questions we got here. Uh perfect says, "Why do you make Do you even sleep?" Yes, so I I mean, literally, I've been on this live stream for what? Like 63 minutes, right? So, 63 minutes to help people learn AI automation and grow. That's 100% what I'm doing, right? I'm I'm just here to help people and and to be useful, right? Uh the reason that we have 145 pages of testimonials from people who have got all these awesome wins like you can see right here. Loads of amazing wins from people whose lives have changed. The reason that we have that is because I create content. And if it only takes me 1 hour on a quick video like this, I will do that every day of the week, right? And so, my mission is to help people learn AI automation and grow with it and to stop people falling behind because that's a serious um situation. And that's my mission, just to help as many people as I can, right? That's why we've got all these testimonials, all these awesome wins, all these people's lives that we've changed, and that's why I create content. And as for do I sleep? Yes, last night I slept 9 hours. What else we got here? What would Gemma 4 be good for? What's the context size? So, anything that you do with AI, it'll be good for. If you don't have a good setup, then you can just use Gemma 4 for sub agents. So, it's very good for sub agents. The context size depends on the model. So, if you download the two smaller versions of Gemma 4, then it's I think 128k context, which is quite small. If you download the bigger models, one of them being 18 GB and the other one being 20 GB, then the context size is 256k, right? So, if the context size is 256k, that's actually bigger than a lot of frontier models like Minimax M2.7. And so, there's two reasons you would use it, two or three, right? Number one, it's free. Number two, you can use it for sub agents. And number three, it has a bigger token context window than most of the big models out there. Can you run tools with the free Ollama models? A lot of the new Ollama models are designed to be run with tools and to run agentically. So, if you look at Minimax 2.7, that is an agentic model. It actually built itself. It's self-improving. It's designed to run tools. I run Open Claw with Minimax 2.7, and you can run that with Ollama. So, yes. So, thanks so much for watching. Appreciate it as always. I will see everyone on the next one. Cheers. Bye-bye.