Claude Agent View provides a unified interface for managing multiple AI agent sessions simultaneously, while Hermes MCP (Model Context Protocol) enables seamless integration between autonomous AI agents like Hermes and coding platforms like Codeex, allowing agents to access external tools, execute tasks autonomously, and work in parallel without manual intervention. This combination transforms AI from a passive assistant into an active automation system capable of handling complex, multi-step workflows such as SEO content generation, website building, and task orchestration.
Installez notre extension pour rechercher instantanément dans n'importe quelle vidéo
NEW Claude Agent View + Hermes MCP Just Changed AI ForeverAjouté :
Today we're going to be running through exactly how to set up codeex with Hermes agents. So how basically how to use Hermes agent with codeex. You get the benefits of both right? You get a super powerful AI super app aka codeex. You also get Hermes agent which is probably the most powerful autonomous AI agent in the world. Right now there's a simple way and a complicated way to do this.
I'm going to show you the the simplest way first of all. And the cool thing about this is like it is so incredibly easy to do. And literally once you've set up Hermes agent, so just make sure you have Hermes agent installed, if you don't have it already installed, you can just go to Hermes agent, follow the setup instructions here. And then once you've done that, you can actually just go to your terminal inside Hermes, inside codeex, right? So let's say for example, we've got like a AI automation blog that we're working on over here, right? You can see it as an example over here. We we can review it, we can check it out, blah blah blah, right? Well, what we can actually do over here is we can click on the terminal option which you can see here. So you can toggle the terminal as you can see right and then the way that you can just run Hermes agent with codeex directly if you want them to run side by side or you want for example codeex to operate Hermes agent here's an example right so you can just run Hermes agent inside the chat we can just check are you working and we now have Hermes agent running inside codeex pretty easy and simple very very simple in fact and the cool thing about this is it has access to this directory. So if you look in the terminal here, it's like here's my Mac Studio. Um here's the folder that it's operating in which is this project that Codeex has built. And then it can use the power of both, right? For example, like you could create a page inside codeex like you can see and then we could actually get Hermes to deploy because it has access to to terminal. Or for example, another benefit could be like right well I've already got a skill inside Hermes on how to do AIO. So what Hermes can do is publish content inside this project folder and we can just set up a schedule here. Let me show an example. So we could say okay take the project that we currently have and publish it to Netlefi as a deployed subdomain.
Right? And then what it's going to do it's going to take all the example pages that we have right here and then use Hermes agent with Netlefi with the skill that's already built in to start using this. So you kind of get the both best of both worlds. Now whilst that's doing that in the background, let me show you the more complicated way but pretty powerful as well. Okay, so what you can actually do is you can go over to Hermes agent and they've got this documentation on MCP. Now what this means essentially is like Hermes agent, it can use MCPS, right? So you could create an MCP and then connect other MCPs, other tools to Hermes agent. It's kind of like a bridge for them to communicate, right? So this is an an easy way to set up MCPS with Hermes Asian, but it works the other way as well. Okay, now you can see here that it's asking me to log into our Netify account. So if I just log in here, it's going to set this up. So I can log in and set that up and verify on the right hand side. So you can see but in the meantime what we can do over here is we can actually set up MCP support inside Hermes agent and link that to codeex and then what that means essentially is like codeex gets access to all the tools that you have inside Hermes agent and it can communicate with Hermes agent directly.
So we can pick a name here. I'm going to go with AI profit boardroom just as an example. Right. So it's asking me for an example to deploy the netifi domain. as you can see and then we can start going from here. Now the other cool thing about this is we could use like chat images. We could use the power of codecs. We could have Hermes and CEX working on the same project at the same time as parallel agents. There's a lot of benefits to this. And if you really love codecs, but you want some of the benefits of something like Hermes agent, this is a cool way to do it. So how do we set up the MCP support? Well, here's what we can do. We can go into a new chat inside codeex as you can see here. And then I'm just going to add a new project. And I'll call this like Hermes codeex MCP.
Right. We'll open that folder up here.
And then we're going to plug in the documentation on the MCP server with Hermes. And what we're also going to do is we're going to grab the documentation from Hermes main GitHub. Right. We're going to plug both of those in to codeex as you can see right here. Right now, what we're going to do after that is set up Hermes MCP with codeex.
So I can use Hermes agent as an MCP inside codeex. Right? And that's the second way to set up. So you got two different options here. One simple, one fast. Right?
one is the and you can see here it's actually deployed the project from codeex as you can see right here right the other benefit of this as well is like if you run out of limits or you hit the token limit for usage inside codeex well the cool thing about that is you don't have to like just stop working on it you can just run Hermes agent inside codeex like you can see here now you can see it's actually deployed this website to a subdomain with netlify this is a project from codeex and So, it's a pretty cool way to to use both and the power of both and just get them working together. All right. Now, if we have a look at a Hermes MCP, that's going to take a bit longer. So, you can see it's saying I'm wiring I'll wire this into codeex as a local MCP server, then verify the Hermes command path and config shape. So, the setup is actually usable rather than just written down, right? And then it's going to start setting up codeex and running codeex MCP with Hermes right there. Right? So two different options which is pretty cool.
You can also do the same thing if you're wondering inside Claude. So I did this yesterday and inside Claude for example you can set up an MCP with Hermes. So you can see that right here. And basically what this means is like Claude can command and control Hermes agent. So we don't have to go inside the terminal.
And we also get the full benefits of Her Claude being the brain and then Hermes agent being the autonomous agent that just goes off and does stuff, right? And again, that's good for limit usage. You also get the nice UI inside cord desktop or inside codeex. And it's pretty easy and simple, right? So now you can see here it says done. Hermes is now registered as global codeex MCP server.
Verified it works over here. As you can see, also made backup to and Hermes gateway is already loaded. So send operations should do right. So, I'm going to say, can you do a test run? We might have to restart codeex to get this working.
And you can see that's now running the session, right? And so it says I've initialized Hermes MCP serve listed the MCP tools then called conversations list and it returned all the conversations and what we've previously done right which is really really cool. So two different ways to use MCP agents with Hermes agent. This is really like a a complete AI power stack, right? So with this you have four different layers. You got Heromy's agent, which is the brain, right? Heromy's agent is a smart brain in the middle of everything.
So, it can talk to other tools, read files, write code, send messages, and take actions. So, it's not just answering questions, actually doing things for you. It's like hiring a really smart employee, right? Who who never sleeps and works 24/7. That's a difference, right? And then you've also got the hands, which is the MCP model context protocol. So, MCP is like giving your AI brain a pair of hands. about MCP. Your AI can't think um or it can think but it it can't really do that much itself, right? It's a bridge between your AI's brain and the real world. And then you've got the builder, which is Codeex, OpenAI's coding agent, right? So CEX is a coding agent that could read, write, and fix code automatically for you. And when you connect Hermes agent to Codeex via MCP, Codex suddenly gets superpowers, right?
because codeex can now do everything that Hermes agent can do um without the sort of guardrails of of codeex and the app itself right and also this way you can use different APIs together right so if we go inside terminal here we can run Hermes like this right and what that means is we can use for example like step 3.5 flash for free is a free API and we've got the full power of GPT 5.5 so like GPT 5.5 is the brain and then step 3.5 flash is the AI that goes off and does stuff, right? The the one the API that doesn't have any limits that you can use for free and then just goes off and gets things done. And then you have the output, right? So when all three layers are working together, you get a machine that runs itself. You know, you can automate SEO, content research, outreach, code, etc. So pretty powerful way to combine them both. Some people say, "Well, I'm not a coder. This is too technical for But as you've seen, if you can copy and paste text, this works as you've seen, right?
Like all I did was paste in the documentation and give codeex access to my local files and you can just talk to it like a, you know, like texting a friend basically.
Also, some people say, I'll set this up later when I have more time. I think that's a limiting belief because every day you wait is a day like your competitors are pulling forever ahead, right? Like the people who set up AI agent stacks in 2024 already miles ahead. the people setting them up right now in 2026 are catching up fast and the people who wait until later will be the ones who are asking for the help of of the people who moved early to help them.
Right? So time in AI is not like time in AI in other industries. Right? Six months in AI is like five years in any other space.
Christo says your message is not supposed to be the agent that's coordinating a codeex a man who was writing big chunks of code.
Yeah, that's it. So like you know Hermes goes off and does stuff. Codeex can write the code, but bear in mind like they can both work together as well.
Like Hermes can write pretty good code, too, depending on what API you have inside there. Uh other people say, well, you know, what's the difference between Hermes and other like stuff like Chat GPT? Well, Hermes agent is fully autonomous, right? It's an AI that can really take actions and automate stuff.
So, you know, if you look at chat GPT, it just answers questions. If you look at Hermes agent, it can send messages, it can read files, it can write code, it can deploy websites, it can do stuff on a schedule. The other cool thing about this as well is like Hermes runs in the background whereas to get tasks done working 24/7 with codeex codeex needs to be open right and so that's kind of a limitation whereas Hermes can run 24/7 whether you have it uh whether you have your laptop switch on or not which is pretty cool as well.
So thanks very much for watching. Just to recap, here's everything you learned, right?
So, what Hermes agent is, Hermes agent is an AI that actually implements stuff, not just answers questions. What MCP is?
MCP is a universal bridge that lets AI agents talk to external tools. Without it, AI is isolated with AI. With it, AI is connected to everything with MCP, right? What is Codeex? Codeex is OpenAI's coding agent. on its own it can do some stuff but connected via Hermes MCP it becomes a full communication and automation agent and to connect them you can just copy and paste the documentation from the GitHub and also the news research research uh notes or to connect them you can just run them inside terminal like you've seen down here right and here's what this unlocks This unlocks all your most powerful tools, right? And real time awareness, plus an agent that runs 24/7 with codeex directly. So, if you want the full guide with a road map, a full setup plan, a 30-day road map, plus 100 prompts on how to use this stuff, I want to put it inside the classroom of the AI profit boardroom link in the comments description or go to the profitboard.com. This is my AI automation community that helps you save time and scale with AI automation.
We've got 100 prompts in there, the 30-day road map, the full framework that I've talked about today, and we've also got guides on like Hermes and Claude, MTP, Paperclip, and Hermes. Everything you need to know about Hermes agent codecs, everything like that we've got inside the AI profit boarding link in the comments description or go to the profitboarding.com. Inside the community, you can ask questions, get help and support. I answer questions inside the communities daily with videos as well. You can get access to all of my best trainings inside the classroom.
Inside the calendar, you get weekly coaching calls. You can jump a full weekly coaching calls, get help and support in real time. And inside the map, you can actually connect with people in your local city who are using AI agents just like you. And if you're wondering like, does this actually help people? You can actually see we've got over 156 pages of testimonials and wins just like this, right? So, loads of people winning and learning and growing with AR automation and just an awesome community of great people who are having a good time learning AR automation at the same time.
So, check it out. AR Profit Boom. Hope to see you inside there.
any recommendations on how I can clean up my disc space with AI? I've started and not completed. Oh, yeah. Yeah. So, this is my favorite way to do it. I actually did this yesterday. Right. So, what you do is you go over to Claude here. I would recommend Claude for this.
seems to be the smartest and the best for this, right? And you go over to Claude and you just like, right, I need to clean up some disc space, analyze my whole drive, and figure out the best places to delete stuff. Give me a breakdown.
Higher space.
start with the options that take up the most space, right? And that's basically how you can do it. And then what it will do is it will run disk analysis, right? It might take like a minute. It will figure out, okay, how much space you have available?
What's the total size? And then with Claude, it will actually figure out, okay, which parts can you delete? And usually what it will do is like give you a table of all the options. Now, one thing to be aware of here is sometimes it doesn't analyze it deep enough, right? So, if you don't get the options that you want or you see like it's just giving you an option to delete small files, tell it to dig deeper and find the files that are taking up the most space, right? Cuz you don't want to be focusing on the files that that have like the least space, if that makes sense, right? You you just want to you want to delete something that's like 100 gig, not like 2 gig, if that makes sense.
So you can see for example here it looks at okay here's a quick win summary right so for example you could delete some of these files this is what's taking up the most space and then it'll give you a breakdown right here so super useful We got a question that says Hermes 3.6 plus is no longer free. So actually did a video about that yesterday. You can use step fun instead if you want to use a free API with Hermes agent. So for example, if we go to Hermes here, you see we got step 3.5 flash and that's running with Hermes setup. So if you go to Hermes setup, then you go to news portal, right? You'll see the models here and the available free model right now is step 3.5 flash. So you can just switch to that. And then if you want to select that, you just hit enter, run through the rest of the onboarding, and after that, restart Hermes and it'll be running with uh step three flash. So that's the free API you can use for this.
Today we're going to be comparing Hermes Workspace versus Hermes Web UI to see which one is the best option. Right? So these are two different ways of running a nice user interface for Hermes agent.
Right? And we can compare them side by side here. Now if we have a look, Hermes workspace and Hermes web UI are both free open-source projects, right? So you can switch between these and you can test them out both. I like to use different extensions with Hermes agent.
I like to test new things, see what's the most useful. And we'll be comparing these side by side to see which one is the best today to use with Hermes agent.
Now some people say, well, why don't you just use Hermes inside terminal like this? Right? If we actually go to Hermes inside terminal, it's it's way sometimes it's way too simple with all the cool stuff you can do with Hermes agent. So for example like if you have a look at Hermes workspace we open this up we can actually set up mobile access directly from Hermes workspace which is pretty cool but then also we have something called Hermes world which is like this cool little AI agent game that we can play and we can just roam around this world and start doing cool stuff right and so there's a big benefit to using these workspaces instead of using like the the default terminal user interface and I think at some point Hermes agent themselves will probably release something that's very similar here um where you can actually just control it with a nice dashboard kind of like open quer own gateway. So what are the differences? Well, let me pull these up side by side. So this is what it looks like when you're using like the default version inside both. Right? So you've got a chat option here for both of them as you can see.
Now, one thing you'll notice straight away is like if we message Hermes web UI here and we just say, "Are you We'll check test this out. See if it's working here. You can see that sometimes they're both a little bit buggy in terms of connecting directly to Hermes. Right.
So, let's just make sure these are working here.
See, like it came up with an error message a second ago, but now it's actually working. So, that's good.
Right. We said, "Are you working?" And it says, "Yeah, I'm here. I'm working.
What do you need?" Right? Now, you can see here it says that it needs to be authenticated, but actually it's replying inside the chat. So it is a little bit buggy sometimes. Hermes workspace especially I found just for connecting directly to your Hermes agent right as you can see right here it seems to be almost duplicated in terms of of linking them together which is not great.
Whereas if we have a look at Hermes workspace, it does tend to work uh sorry, Hermes web UI, it tends to work straight off the bat. Now, if you need to fix this at any point, what you can do here is you can just paste in the error and just get them to sync like this.
So you see all the news any called auto use will be overcharged. I haven't seen that actually.
Thanks. Just changed to step fun. Happy to help.
Now you can see that's fixed here. Right now there are some pretty cool things inside both of them that you can use. So let me show you side by side how that works. Right. So if we pull up Hermes workspace, what you see here is a chat. So you can chat with your agent here. And then you've also got something called swarms.
I think this is probably the best feature inside Hermes workspace. So what this means, you can basically have like 17 different profiles of Hermes working at the same time, right? And so that's the only way I've seen to really run a full agent team apart from Ruffalo with Claude for creating like a full agent swarm, right? Like for example, they built out a 50-page website for me. If we have a look at, for example, something like Hermes web UI, you can see here there's no swarms option, right? We have the Hermes dashboard. We can see all the profiles that we've got, but you can only use one profile at a time and you switch between them like that, right? So, you can choose which one you want to use and which one you want to start using on, right? But you can't use them all at the same time. So, if you want like a big theme of agents, Hermes workspace is pretty good, but just bear in mind like it is a little bit buggy. Like even looking at this, this is the older version of Hermes Workspace.
But I've just pulled in the latest version, right? So, it's kind of like it's a bit weird sometimes with the configuration and getting them to work.
Usually, if I'm having problems, I will get Hermes or Claude to just go directly and go and fix that. The other thing that's pretty cool here inside both of them is you have this spaces section, right? So, you can manage all of your files. So, for example, if you go to Hermes workspace, it's actually easier to manage all of your files because you'll see them all inside this section.
you can edit them, change them, um move them however you want and that sort of thing. You also get a terminal option here. So if you do prefer to use like the Hermes web UI, you can use the terminal whereas I think Hermes web UI itself versus workspace is a bit less customizable, right? It doesn't go in the same sort of like granular depth that you normally get inside uh Hermes workspace.
Now, there's some other customizations that you get inside her workspace. You can see them here. Like, for example, you've got the camb board. So, you can actually add like new tasks, assign them to an agent over here, and then you can get them working for a camb board, which is pretty cool. You can set up jobs.
you've got a conductor. So this like an office as you can see and then you can actually launch a new mission for all your agents to work towards inside Hermes workspace, right? Whereas you you don't really get that inside something like Hermes web UI. So I think if you want to manage like your agents one by one and you want something that's very simple that connects pretty much perfectly every single time and doesn't require um anything too complex then I would go with Hermes web UI right whereas but if you want something like with all the bells and whistles all the crazy stuff all the the really interesting things you can do with agents then you can go with something like Hermes workspace right so you've got like for example the operations, you got the swarms, you have the camb board, the conductor, etc. Right?
There's a lot more customization that you can do with Hermes workspace. So that they are the main differences between them. Um, just be aware that it can be quite buggy with Hermes Workspace as you've seen today. Also, you don't get the game inside Hermes web UI whereas you do get Hermes world inside Hermes Workspace. So those are the main differences here.
Now, if you want a full guide to using both, for example, Hermes workspace, we have a full setup guide with a 30-day road map, 100 prompts, and a step-by-step plan on how to use and implement this stuff, as you can see with Hermes a uh workspace right here.
And then if you want to learn how to use a web UI, we have that as well inside the air profit boardroom. So pretty much everything you need to know about Hermes agent, how to use it, all our step-by-step guides. They're updated daily with new cool stuff as you can see inside the air profit boardroom. link in the comments description or go to the AI profit boardroom. This is my community that helps you learn, grow, and scale with AI automation. Inside the classroom, you get all my new trainings.
Inside the calendar, you can drop a weekly coaching calls. You can connect with me personally and it's got everything you need to learn about AI automation and AI agents in general.
Let's see what questions we got here.
So, Abdul says, "Please application name." Yeah. So, Hermes workspace and Hermes Web UI both available on GitHub.
Trad says hopefully tokens token costs go down. So you can actually use Hermes for free, right? If you want to use it for free, then you run Hermes with news portal and you make sure that you have step fun 3.5 flash setup right there. Right. And that's how you can use Hermes agent for free.
Second uh says, "Please talk about how people with less than 32 GB of RAM and no GPU and want 90% local or if you if you want local, you need a good setup, right? So if you've got no RAM, you've not got a good setup. You wouldn't run a local setup, right? You can just use API instead. For example, like G uh step 3.5 flash."
And then he second says, I'm using Docker and Telegram to talk my Herb agent. I want it lean and max efficiency. Uh I don't actually use Windows setups. Right. So I know you said there, can you talk about Windows setups? I don't use Windows at all.
Simply just I don't find it good for productivity and I prefer using a Mac Studio or Mac Mini if I'm going to run local models at all. So those are the differences.
Today we're going to be looking at the new Claude code agent view which means you get one list of all your sessions and it's available as a research preview. Right. So you can run clawed agents to start dispatching multiple sessions at once, right? This is pretty cool because you don't need like a terminal tab running to run this. You can just have a agent view and you can have these running together as separate sessions, but you can easily manage them with this. Right? So when you're running like agents in parallel before, you have to like manage multiple terminal tabs or have like a big grid of of terminal agents, right? It just got super messy, right? Whereas for example, if you're using agent view in claude code, then you can manage everything at once and it's much easier to organize, right? So let's talk about how to use it. So you can press the left arrow from any session or you can run claude agents from the terminal to open agent view.
Let's try this out. So this is inside the terminal, not inside the desktop app.
Let's test this out now.
and then we can start using this. Let's try this out.
We're going to create multiple sessions here.
and then we can start a task in the background. Right? So once we've typed in claude agent, we can then start running a task in the background. So if we say okay, keyword research for AIO, that's going to start working here, right? And so we can have multiple cla code sessions running in the background rather than having separate terminals.
So we can have separate terminals like this. But this gets quite messy, right?
Sometimes you lose track. You don't really know what the name of these are, etc. You don't really know what it's working on. Sometimes you forget to check it, etc. Now you can just have all of them running inside one single terminal window. So we've got this working on keyword research. We can have another one creating a website. So let's say create a simple HTML landing page.
And now that adds it to the working list, right? And so we've got multiple different Claude code agents working in parallel. We can see them with the agent view and we can see what's going on.
Right? We can set up another one here.
So create an AI SEO calculator for organic keywords.
And now it's working on those three. And you can click on it or uncclick on it.
So if you want the full view of what's going on, you just click on them like that. Right?
And then you press left to go back. So if we click on this one, these are three different tasks. As you can see, if we click on this keyword research task, we can go straight into that terminal for it. And we can see what's going on.
Right now, if we press left, then we can go back here and we can also see now that one task was completed. So this website was completed all from the agent view. And so like it's really crazy because you can deploy multiple tasks at the same time. You can be like way more productive, get more way get way more done and it's all running in the background whilst I talk to you, right?
And we can click on that and then we can go back to the task here. And so the way that you manage agents is really changed forever because instead of having like multiple terminal tasks or for example having multiple tabs or for example setting up a new clawed session in each tab like that which is super messy and difficult to manage. You can have them all running in one view which is super useful. And then you can peak and reply without leaving.
So you can select a session to peek at the last turn. If it was waiting on decision, you can click on it like that.
You can press left to go back and it's super easy, right? And then if you want to see what's completed, you can see an example right here. It's really, really cool. Now, this is something we're using auto mode as well to get stuff done. So you can see here we've got auto mode running, which just means like the agent can work autonomously and make decisions autonomously without me handling anything. And then you can see the task is ready to go here. So if we open this up, we've just created a a HTML page just whilst I was talking to you. So that is how agents have really changed because you know previously you would have one by one or you'd be managing like five different terminal tabs across your screen. Not a great way to manage your agents. Now you can manage them just in the one uh simple view like that. It's way easier. Now you might say okay why would you use this? Well in terms of this number one you can scale the number of concurrent sessions. So you could dispatch several tasks and projects at the same time, each one paired with a skill and then return to a list of pull requests ready for review.
That's an example, right? You could manage longunning agents. So you could actually have like agents managing your other agents in the background here and you can easily navigate between separate sessions which is just more productive, more easier, right? And you can see what's actually completed. So with the completed status here, we can see exactly what's happened and we can check them out and that sort of thing. So really simple improvement, but it helps a lot and it just makes it much easier to manage your agents in one place.
So thanks so much for watching. That is how to use the Claude code agent view the new setup from Claude Code agents.
If you want to get more training on Claude and how to use it, etc., You go to the air profit boardroom link in the comments in description or go to a profitable.com type in claude and you can see we've got loads of different trainings on for example like claw manage agents claw code persistent memory how to use claw code skills etc. And we have full tutorials and step-by-step guides on everything that you can see today. So if you really want to learn and master this stuff we got loads of training inside the air prof you can ask questions get help and support etc. I create daily videos helping people as well. Inside the classroom, you can get all of my new trainings and setup uh including new daily advanced tutorials right here. So, every day we just adding new advanced tutorials and showing you exactly what works, etc. with video guides and step-by-step tutorials plus 100 prompts on how to use each one. And you can see all the trainings inside the classroom here. So, you get loads of cool stuff inside the calendar. You can jump on four weekly live coaching sessions. So you can jump on these live AR automation sessions, ask questions, get help and support, share your screen, meet other members who are doing similar things to you and inside the map you can connect with other people who are using AR automation and Claude to help you grow and scale, right? It's just an awesome network of people who are doing amazing things. So feel free to check that out. Plus you can connect with me personally inside there too.
Agents managing other agents like sessions connected. Yes, absolutely.
Right. So if you go inside the terminal, you can have Claude agent view here managing your sessions. Now we could create a new task. It's like, okay, uh, set up a new skill inside Hermes or AI SEO, right? And then that can just manage my other agents in the background. So, we can go over here and it can start managing our Hermes agent, speaking to it directly using the Hermes agent MCP, and then we're good to go. Right? I'm actually going to stop that because I don't want it to do that, but I just wanted to show you an example of how that works. And then it'll be completed once done.
And you just set up Claude agents to go back to it.
What's the difference between Claude and Codeex? I mean, like if you look at Claude, right, it's more It depends what you're comparing, right?
If you're comparing like the terminal usage like this, honestly, like Claude I find is like a lot more useful for getting stuff done and being a lot more autonomous. Codeex is good. It tends to run out of credits pretty quickly. I found even when I'm on the pay plan. So, that's one thing to be aware of. And also, the Codeex UI is totally different, right? It's totally different. So, there's some big differences there. They do similar things, but it just depends like what's your favorite API. For me personally, I like to test and use them all because I like to learn about this stuff.
What's the difference between Codex and Claude and Hermes agent? So Hermes agent is open source whereas Codex and Claude are not open source, right? Which means that you can customize Hermes agent as much as you want to. And also you're not stuck to one API with Claud code and code and Hermes agent, right? With Hermes agent, you can switch between APIs. It's a lot more autonomous. It can connect to your Telegram and everything else. You can't really do that with Claude. You can connect it to your phone, but it it's not very good the the phone connection, honestly. And also, you just can't customize Claude in the same way, right? And also, you know, Claude Code is is paid, right? Whereas for example, if you're using something like Hermes agent, you can use it for free because it's an open source free project and then you can get a free API to plug into it like step 3.5 flash.
Is there a way to allow Google uh Hermes to log into Google? Yeah. So you can basically just give your login details to Hermes agent and then it can use that email address. The other thing you can do is like you can go to Google Workspace, get an API and then give that API with the um authentication allowances to to use Hermes agent. So Hermes agent can connect to email that way.
Is this the real Julian Gold? It is indeed.
Brian says, "I would love a handover loop." So when context is maxed, the agent does an automatic handover and the agents keep working with a new context.
So you can just um comp usually like Claude will compact automatically if you need uh if you run out of context, right? So you can just for example, you can go into uh Claude desktop like this and you just type for/compact, right? So if it ever runs out of um context, which I've never really seen, I don't really see very often unless it's a huge huge uh task, but if you're just inside claude desktop or inside cloud code, you can run compact and then that um basically makes the context window smaller so it compacts into a small window. So that's the way that would do it.
How can I use Hermes agent to work on my mobile app and how can I work with Hermes agent for marketing and so on? So for me personally like I can use Hermes agent for like editing videos for creating SEO content for doing daily blog content etc social media content.
So that those are some of the ways you can use it for marketing. We have loads of training inside the app offer board on that. And then when it comes to getting Hermes to work on your mobile app, you just go into Hermes setup and link them together. Right? So if you go into Hermes setup like this, we just enter on this section.
And then you go to the select platforms to configure section here. Right? And that's how you can configure your channels that you want to use with Hermes agent. So if you want to link it to Telegram, just configure it with Telegram right there.
Oh yeah, I'm doing a huge task. Compact can lose some details. So I usually get a handover document. Yeah. So what you can do is you can get um you can get all your agents to type any key notes as they go along into something like Obsidian, right? Obsidian is basically like a second brain. So if you need to store context and notes, you can use obsidian with your AI agents and it'll write down the notes inside a markdown file so that you never lose context again. So if you're finding like you're losing key details as your agents go along, for example, like Claude, what you can do is tell it as it works to note everything into Obsidian. Write down the notes inside Obsidian and then you've got all the keyyn notes noted down as it works, right? And then also all your other agents can connect to Obsidian at the same time so that any context you have inside Claude gets automatically transferred to Hermes or Codeex or whatever you want.
Isn't Claude dropping third party configurations?
Uh, it depends what you mean by that. I don't know exactly what you mean, but I think they they've dropped some stuff, but like you can still connect easily to uh different things with the customize option. Right. So if you go into claude here and then you go to your settings like this go to connectors and you can see they've moved it to customize and that's how you can connect to different MCPS right so for example we've got like Hermes MCP connected there WordPress GitHub cla chrome etc right that's how you can connect to third party applications is Okan says, "So Hermes agent is better for coding and front end design than Codex Claude." Well, I mean like they're all pretty good at that, right? It just depends what you want to do. Like you can use any of them really. It depends what you prefer. Like if you prefer to manage AI agents from your phone, then Hermes agent is going to be better for coding and front end, right? If you're on your laptop all day and you prefer the UI of Codeex and GPT5.5's API, then you can switch to that. So, it really depends what you want to do. For me personally, I think the best for front end is Claude, but Hermes can do a pretty good job, too.
So today we are going to be looking at a brand new update inside Claude which is the goal completion right and what this means basically is like you can set up a condition completion condition with goal and Claude keeps working across turns until the condition is met. Right? So the forward slash goal command basically sets up a completion condition and Claude will keep working towards that goal without you prompting it after each step. Right? And after each term a small fast model basically checks whether it's actually set up, right? If it doesn't, Claude just keeps going until it's completed, which is super powerful, right? So basically like you can say okay keep going keep working autonomously until you complete this task. You can also use loop as well. So there's two options here. You can use goal and you can use loop. Loop continues when the previous term finishes and it stops when the conditions are met and judged as completed. a loop. Basically, you can set up a time interval elapse and it stops when either you stop it or when Claude decides it's done. Right? There's two different ways to make Claude way more autonomous as a kind of AI super agent using this process. So, let me show you an example of how this works.
And we're going to be using this for AI SEO, right? So, I'm going to use this to show you how you could, for example, like rank and build a website. So we can say for example for/goal build a complete SEO optimized website structure for the AR profit boarding and then we can give it a really complex front. We can say for example okay the site must include a homepage with the AR profit boarding details a hero section with the benefit bullet points social proof section a single CTA and then we also need uh five core landing pages each with unique H1s 300word introduction H2 sections and an FAQ. Then we could also use a blog structure with 20 article titles, targeting keywords like AICO tools etc. A full internal link map connecting all the pages, schema markup suggestions for homepage, landing pages and blog posts. And every page must have a unique meta title under 60 characters and meta description under 155 characters. Right? And then we say don't stop until every page, every article outline, every meta tag, every um internal link has been completed and documented. Right? So we can type that in like so. Make sure the website is deployed locally as HTML.
Plus, it's not complete until the website is fully ready, right?
And so what this means is you can make your AI agent way more autonomous. It's just going to keep working towards that goal and get it done until it's completed as you can see right here.
So you can see that's generating it now and thinking about the task.
And basically you might say, well, why is this so useful? Well, every requirement is verifiable, right? Claude literally can't stop until every single output is surfaced in the conversation.
So the evaluator model checks each condition after every turn. If the internal linking map is missing, it keeps going, right? If a meta description is over 155 characters, it's going to fix it. And so you can walk away and come back to like a full website architecture ready to hand to a developer or drop straight into WordPress or Netify or whatever you want, right? It becomes very very autonomous as a powerful AI agent.
And so this is a really powerful way to to rank your website. So let me show you an example of like websites we use AI SEO for that have been built with AI. So let me pull up some examples here on Google search console. I'll show you some of the results.
So you can see for example this website is is growing massively in trajectory and it's improving all the time. Right?
Here's another site that's growing massively in traffic. I mean this was basically at two clicks a day. Now it's over 100 clicks a day.
This one was at six clicks a day. Now it's 69 clicks a day. Right? And here's another one. So this one was at what like 20 clicks a day, something like that.
Now it's at 500 clicks a day using these sort of processes. And so you can see here it says um okay I'm going to go off start creating this go off and build it etc. One second.
I think it And now it's actually spawning multiple websites to work in parallel on all the HTML pages.
All right. So I can just walk off, come back. It's still working on the task and it will just keep going until it's done.
Now this is terbased. What that means essentially is that it will submit a first draft. If that draft doesn't meet the completion criteria, which is judged by an evaluator model, then the judge is like, "Nah, no, no, that's not ready to go, mate." Right?
Like, you've got to go off and do that again. And so, it just keeps going and going and going until eventually it's done. That's the biggest difference here. And it's it's really a huge difference in how you operate AI agents.
Bear in mind as well, like you've got the agent view with Claude as well. So, it's pretty powerful. Now, if you've already submit the task and then you put forward/goal after it says no goal set, right? So, you have to put forward/goal first and then the prompt that you want to build like you can see right here to get it working.
You could also, for example, just take like all of your keywords you want to rank, like we could paste all of those in and say like, okay, create the content based on my criteria. Do not stop until that is completed, right? Um, and also the cool thing about this is like Claude agent itself, you know, Claude code can spawn multiple agents in parallel to build more content. So, it's like you get a full team of agents. You got the evaluator agent that's checking if the job is done. You've got the goal agent that's actually doing the work.
And then you've also got the sub agents that are being spawned in parallel to get this job done. Like it's it's really crazy. The one thing you have to be careful of is like this is going to use a lot of tokens. So, you have to be careful about that. But, it makes it so powerful in terms of how it can be used.
You've also got auto mode which combines with that. Right? So auto mode on its own like approves tool calls but doesn't start a new one. Right? Claude stops when it judges work is done. So goal adds a separate evaluator. So you're combining the power of auto mode inside claw code with the power of goal right to to make sure that job is done until it's completed. That's the difference here. So for example, it's not claudit auto mode won't just keep going on its own like this if the job isn't done right. Whereas for example with goal it will keep going over and over again until the evaluator says yeah this is ready to go. I mean some other examples you could run goal followed by the condition you want satisfied. So for example you could say okay forward slash goal all test in test pass and the lint step is clean right to make sure it's not. Now that's like more of a technical one. This is more like a simple version that you can use that's actually useful for people who are non-developers, right? And you can see it's like ticking off the task at the bottom over here to make sure it's actually getting done. The other thing to note is like if you run another forward/goal in the chat at the same time, it will actually cancel the existing goal. So just be careful of that.
So really you can have SEO running 24/7 using this process. You know, it's not a dream anymore.
Claude's goal command does this, right?
You tell it what you want, it keeps going until it's done, right? You don't need to prompt it every five minutes.
You don't need to hire a team of 10 people to do what AI can, you know, one AI can do.
And I call this the goal SEO stack. This is a custom framework that I recommend for using this stuff, especially if you're using it for SEO, right? But you could use it for anything in your business. So the old way of doing SEO was like hiring humans to guess. The new way is like giving AI a target and just letting it run, right? So most people think SEO means writing a blog post, hoping Google notices, waiting 6 months, blah blah blah. That's the old model, right? The new model with the goal SEO stack is a completely different paradigm because it's not about doing more SEO work. It's about setting a verifiable condition and letting the AI loop until it's achieved. And here's how it works.
So step number one is you want to generate a verifiable target. This is G, right? So you don't just say brand me some SEO content. You say I need 10 fully optimized articles each passing a readability score of 70 with internal links FAQs and schema markup right and goal doesn't stop until all 10 are done right that's a target that claude can verify and that's what makes it autonomous now you've then got the orchestrator section right so claude doesn't just write once it checks its own work after every turn and a separate AI model a fast lightweight model evaluates the condition after each step if it's not done claude keeps going. If it's done, it stops automatically and you're not involved in any of the loop. Before you would like have to orchestrate and manage it yourself. Now you're not even involved, right? You've really got like the evaluator model that's the CEO of this whole process.
And then you can also automate the proof, right? So the genius of gold is that Claude has to prove the condition is met, right? So all 10 articles are written only works if like Claude surfes the output. Keyword density is above 1.5%. For example, if you set that as a target, that only works if Claude runs a check and shows you the result. So you bake proof into the condition, which makes it verifiable and not vague, right? And then you can layer in your SEO SPs, right? So the goal command is the engine, but your SP aka your step-by-step operating procedure for creating content is a fuel. Let me show you an example.
So this is my SOP for creating content.
You know, it's got the keywords, then it's got the content outline, then it's got the content itself, right? And a quality control checklist. So what we could actually do here is we could turn this whole SAP into a goal-based process. So I can paste that into Claude and I say, okay, create a goal-based prompt for this, right? Based on that SOP that we gave it. And it's like, okay, here's what you're going to do. You're going to implement all four phases and don't stop until every output is complete, right?
The keyword research, the content outline, the content creation, the quality control, etc., right? And so it has to complete all four phases before it can even think about submitting this.
And then if it's not right, it will just keep going until it's done. Right? So it has its own kind of like quality control process without a human in the loop, which is pretty insane.
Now, some people going to say, "Okay, AI content doesn't rank on Google."
Honestly, AI assisted SEO content optimized with the right structure, internal links, and internal intent matching ranks all the time, right? Like loads of people in the AI are profitable boarding are SEOs and they're proving this right now. And I've shown you proof on my own websites, too. Like you can see, so this process actually works.
Like the belief that AI content doesn't rank comes from 2022, but it's it's 2026 now, right? Google ranks helpful content. AI can produce helpful content at scale. So the game has totally changed in terms of that.
Cory says, "What do you think of generative engine optimization as priority oppos?" Honestly, I see them all under the same umbrella, right? AO, GEO, SEO, it's all the same thing. It's how can you rank inside AI and how can you rank inside Google. So this same process I'm showing you today, right?
And all the processes that I talk about in my videos actually work for us, right? And you can see us ranking on Google multiple times here, right? You know, you've got a Reddit post over here, you got a website, you have um some of our content here as well. You have another website over here, right?
And so we can rank inside Google. Now, that same method can rank inside Google AI search engines as well, right? So you can see us ranking here and you can see us ranking multiple times down here as well, right? And so the same process that you use for SEO works with GEO.
It's all the same thing, right? There's no difference between them. Um the main thing is you know the question you want to ask yourself is not are you doing generative engine optimization or SEO. It's are you ranking right? Are you getting traffic from search engines? And you can see the process I'm showing you right now is working. So however people want to label it, you can. But the main thing is you got to focus on like are you ranking or not? not are you doing go or ao whatever right other people say well I need a big team to do SEO at scale but honestly the right belief here is like one person with claude the goal command and a clear SOP can produce this 10 person content team right with one loop one condition 10 articles boom shackleaku it's done right you can see it building out the website right here other people say well I'm not technical enough to use AI for SEO but if you can type a sentence you can run goal there's no coding there's no setup there's uh beyond type and a condition. You write what you want and Claude does the rest as you can see over here. Obviously say well SEO takes months to see results. The right way is like AI lets you produce in days what used to take months, right? Um, and you can collapse like the production time to almost zero if you do this, which means you can rank for 10 times more keywords in the same time window, right? So, for example, if we type in this keyword here on Google, you can see that this content that we produced 24 hours ago is ranking first, right? This is another website page that we've created ranked within 5 days for this keyword. So it doesn't take a long time to rank with SEO if you have the right processes in place which is what I'm showing you today.
Other people say I can just use chat GPT for this. Chat GPT codeex does have goal but chat GPT directly doesn't have a goal command. Right? So chat GPT will stop when you stop prompting. Claude goal loops autonomously until the condition is met. And that's a totally different workflow. Right? is an autonomous AI agent that's working 24/7 towards a goal to help you rank on Google, right? So, super powerful stuff.
So, just to recap how it works. So, you type your goal followed by your condition. Claude starts working immediately. After each turn, a fast AI checks if the conditions are met. If they're not met, Claude keeps going autonomously.
If they're met, it stops and shows you everything it does, right? And you don't have to do anything between steps two and step five. That's the magic for this. So it saves a lot of time. Now there's also other workflows you can use. So there's not just for/goal, right? You can use for/ loop which runs on a timer and keeps going until you stop it, right? You've got for/goal which runs on until a condition is met.
So it stops automatically once done.
You've got auto mode which approves tool usage with a single turn. So you still prompt each turn. And then you also have goal and auto mode together which is the ultimate combo. You can see I've got that running here. Right? So there's no tool approvals, no manual prompting, just pure autonomous implementation. And for SEO work, really the goal feature is the one you want. Right?
So the other thing about this to note is like most SEO work is very repetitive.
You write a brief, you send it to a writer, you wait a week, you review it, you send back edits, you wait another three days, you publish, and you repeat that 50 times. Very, very repetitive.
Even if you're using AI, right? with goal. You type your condition, you walk away, you come back to 10 finish optimized structured articles or whatever you want. And that's the difference between like SEO with AI in 2022 and you know SEO in 2026. It's the old way versus the new way, right? And the people who adopt this now will have an insurmountable content advantage and optimization advantage in 12 months.
Everyone else will still be writing briefs and messing around with that stuff.
Now, we've got loads of prompts for SEO on how to do this, as you can see. So, we've got prompts for local SEO, for ecom, etc. for um analytics and reporting prompts as well, which is super powerful, and a 30-day road map all inside the AI profit. Right? So, if you want to get this full framework with the goal SEO stack, the 100 prompts, the full road map on how to implement it, etc., You can get that inside the AI profit boardroom. If you just go to the AI SEO automation section inside the classroom here, I've added the full section with all the notes as you can see right here. But that's basically it. Now just to recap here and you can see that it's checking if things are done as you can see over here inside clause it's like good 13 articles are done seven are missing I'm writing them all directly here right um so just to recap on this stuff we've talked about the goal SEO stack so you want to generate a condition orchestrate the autonomous loop automate the proof and then layer in your SEO SOPs for doing this right the key takeaways here is like goal is the most powerful SEO tool right now. It loops autonomously until your condition is met and it just keeps going until it's done, right? And also one thing to remember is like every SEO task that involves repetitive checking, reviewing and iterating is a potential goal task, right? So if you find yourself checking something manually more than once, that's a forward/goal. Build the condition, set the loop, and then you can come back when it's done. So, if you want to go deeper on this stuff, inside the air profit warning, you get the full interactive version of this guide, live walkthroughs, weekly AICO updates, direct access to a community of 2,900 agency owners, the complete eight agent SEO prompt pack, and um and you know, you might be wondering, okay, like are people getting results with this stuff?
You can see that we actually have over 156 pages of testimonials, you know, from people winning and learning and growing with this stuff, right? Wait, look at this one. Ash says, "My agent literally runs my pizza shop." Right.
Um, Francisco says, "My blind my mind was blown since I signed up."
Michelle says, "Outstanding." Right?
Like, there's so many wins inside here.
So, if you want to get access to this, you know where to go. Um, inside here as well, you get an amazing community you can ask questions to. Inside the calendar, you can jump a weekly coaching calls. Inside the classroom, you get access to all of my best trainings.
We've got a full AIC automation section here. And inside the calendar, you can jump a weekly coaching course.
Now, also, if you want a free one-to-one SEO strategy session that shows you onetoone how to grow your website, how to get more traffic, how to rank inside Google and AI search engines, you can book that in at goldie. Agency, link in the comments description, and you'll get a free SEO domination plan. So, you get a custom tailored game plan to generate more leads and sales from your website.
You'll discover the secrets SEO link building. You'll be able to ask our world-class experts and about any questions, problems, or challenges you're facing. You'll learn the best link building strategies for your website and how to outrank your competitors with link building. So, that's available at goldie. Agency.
Let's see what questions we've got here.
So, Brian says, "Codeex Windows app gets laggy. I use terminal for it now." Same for me. I prefer to use terminal with clawed code, honestly. Um, unless it's a massive project that I come back to every day.
Brian says, "I use AI around 10 hours every day." Wow, that's amazing. I mean, that's the level we're at. That is insane. I agree with you there, Brian, mate. Thank you.
Cow says, "This is nuts. AI evolution is hitting critical mass. You helped me days ago and things have improved exponentially since then." I'm happy to help. So, thank you so much and good for you Cal for uh sorry, Kai for for taking action there. Well done.
And if you ever want to stop the task, just press control and C inside the chat and then you can stop the whole task.
All right. So, today we're going to be setting up a custom automation for an air profit boarding member. And Julian Boss has said, "Real world example. You know, with all the various AI tools out there, could you create a tutorial on how to build something super useful, for example, like an AI powered personal travel agent, right? So, here's a real world use case. He's planning a trip to Manila. Travel windows between these dates. Wants to stay for exactly 9 days at this place. What's the most practical way to build a travel agent? Right? So, here's how you can do it. And so, if we want to use Hermes agent to help us build this, we can open up the terminal window here.
And we're just going to run this with Hermes agent. By the way, I do this for members every day inside the air profit boarding if you want me to do it for you, too. And what we're going to do from here is we're going to say, "Okay, help me automate this." Right? And we'll plug that in and then we'll get Hermes agent to help us set this up and to start automating this workflow. Right? Now, the first place I will start with this process is I'm using the Hermes agent OS system, right? Which is my framework for building an AI powered agent or process that runs itself, right? And this is the framework behind every automation that I build, every agent that I deploy, every real world use case that I show on this channel, right? And so the first thing you want to do is really look at like, okay, what's, you know, before you build anything, find the task that's using up all your time, right? It could be research, could be outreach, whatever you're spending most your time on, right? You want to find it, name it, and own it, right? Then you're going to engineer the agent. So, you're going to build a Hermes agent with one specific job, right? Like you can see right here.
So, you can give it a goal. For example, you're a travel agent specialist. Um, you can give it a role. Sorry. And then the goal would be like find the cheapest flight, right? You can give it the tools. And now it knows exactly what to do, right? So, you can see it's beginning to build out that project structure over here.
If you have any questions as we go along, by the way, like feel free to ask.
So you can see it's beginning to build this. We could also do this with Claude as well. Claude is another way to set that up.
So we can go inside Claude here and do the same sort of thing. So whatever you prefer, you can roll with it. We can run both tasks inside Claude and Hermes at the same time. Just see what happens.
And then from here, you want to run it in the real world, right? So you deploy the agent on a real task, which is what we're doing over here.
You run it, you watch what happens, and then you catch the gaps, right? So with these sort of things, it's not going to be perfect first time you use it. Just bear in mind with that, but you're going to catch the gaps when you test it out. So done is always better than perfect. And running agent that's 80% right beats a perfect one, still being desired, right? So that's what we're doing with the HES agent OS system that we've got here.
And we're going to set it up as a skill inside Hermes, which means that anytime we use Hermes, you know, whether that's on Telegram or anything like that, we can use it as a skill inside Hermes and then call upon it.
Now, whilst we're waiting for this to be done, the other thing to note here is like you can multiply this with agent stacks, right? So, one agent is good, but you could have a team of agents that's unstoppable and you can connect them together, right? So, you know, agent one could be, for example, like a SEO research agent. Then you could have a separate Hermes agent profile for um creating content. You can have another one for creating avatar videos. You know, you can have multiple of them, right? You can see how Hermes is now creating the skill wrapper for this AI travel agent planner.
Claude is doing the same thing, but it's approaching it in a separate way, right?
So, you can see it's actually building the files in parallel, setting all of this up, as you can see.
And then we can say, you can see here it's actually built the back end of it as you can see right here. Right. Um, we do need to get some APIs for this, but we can do that in a second.
And then you can actually set up like a simple mini app on a website where we can use this. Right. Now for travel related stuff, we just need to get an API key. We can get Amodus API. So let's go off and do that.
And now you can see for example we've got this agent ready to go. Right? So we've set this up. It's like right here's the trip monitor. Here's the details. Then we can put the details in here and just start using that. Right?
Obviously you need to plug in your API key from um from Amadeus to get this working but that's essentially you know how you can get an app running for whatever you want right and you can see it's hosted locally here so no one else has access to it and that is the setup with the trip monitor details and so you can build and automate anything that you want with Hermes agent especially using this operating system that we've talked about and it's pretty simple and then you know the final Fin step is just expand to new cases, right? Once a workflow is running. You know, what else could you automate? What else could an agent handle? What else um could you offer clients or what niches would be interested in this? Right? You can expand, you can duplicate and uh you can grow from there, right? And then you just systemize it, right? The the main thing here as well is like as you go off and you start building SOPs and that sort of thing, you want to systemize it so that you can come back to these SOPs whenever you want to. That's actually one of the reasons that I built out the AR profitable boarding is because everything is like organized and I can just if I need like the SAP for example for um Hermes and computer use I can just type that in here and then get access to to computer use agents um inside here or if I'm like I need to get the SAP for open web UI here it is boom I can just get access and type in the search bar. So whatever you want to build and learn, you can do it using this process. But just make sure you organize it into SPS piece as well, right? And the goal isn't just to automate one task. The goal is to build like an entire operating system for everything um that runs on AI and automates in the background. That is the Hermes agent OS, right? And everything is just built on top of it. Now, some people say, well, I need to learn how to code to build AI agents. As you saw today, it was pretty easy to set up, super fast, right? And Hermes agent just lets you build powerful automations in plain English. So you type what the agent wants to do, what you want the agent to do. It figures out how to do it. No coding required, no Python. Um just clear instructions, right? Also, some people say, well, AI agents are only useful for tech companies and developers. But you know, the biggest wins right now happening in the most ordinary stuff. I mean, I've just shown you how to automate a travel agent. It could be real estate. It could be ecom, could be freelancing, etc. Marketing agencies. Other people say, "Well, it's too complicated to set up." Right? The first working agent, as you've seen today, could be running in 30 minutes or under 30 minutes. So, it's not days, it's not weeks, is like 30 minutes and a good API. That's literally it.
So, if you want to get the full framework of how to use this, um, 100 prompts and a 30-day road map for implementing this into your business, feel free to check this full guide out.
I'll put it inside the AR profitable boardroom classroom over here in the daily update section.
It's called the Hermes agent OS system.
And that's basically how you can automate anything with Hermes agent.
Pretty amazing.
So, I've created that custom automation for Julian. If you want me to do stuff like this for you as well, then uh you know, I do it for free for members inside the air profitable boardroom.
This is just like a little bonus I give members to help them. And uh hopefully that helps you understand how you can automate anything with Hermes and the Agentic OS. Right? This is my AI automation community that helps you save time, grow and learn and scale with AI agents and AI automation. Inside the community, you can ask questions, get help and support. It's a really active community with loads of cool people, lots of cool posts. Um very interesting stuff that people are working on as well. You can go inside the calendar here and jump on four weekly coaching calls. Share your screen, ask questions about your setup, etc. Inside the classroom, you get access to all my trainings, all my best courses, etc. And we have a new daily updates here if you like advanced tutorials like you've seen today. Inside the map, you can connect with members in your local city and area so that you can meet other cool people doing similar things to you. And that's all inside here. Link in the comments description or go to the aircraft.com.
Let's see what questions we got here.
Brian says, "Yes, I'm obsessed with AI.
I don't know if I'm learning this much to you. Thank you though. trying to make a dream come true. Thanks for tips.
Happy to tell. Christopher says, "Hey, that's my floor waiting to ship until perfect." Yeah, I mean, as you've seen today, like that setup will not be perfect. Obviously, you need to connect the APIs as well, but you can see you can see that we can quickly build something, create like a first prototype, and then just test and improve it from there.
I miss says what models are you typically using when creating these systems apps and agents. So Hermes agent is the one that I use for this and then we use that with the free API step 3.5 flash and if ever in doubt ask AI. There you go. Is the digital is this this is the real Julian Goldie? Yes. I'm currently using open router with three models then let letting it switch to auto on its own. Is that one way to build stuff like this with a flexible model? Absolutely.
Yeah, you can switch between APIs like you've seen today.
So, one of the questions I get all the time, as you can see in this example in the profit boarding, is that Richard was asking like how to use Claude as a second brain, right? How, you know, if you wanted to set up Claude as a second brain, how do you do it, right? So, this is my process.
All right. So, I use a combination of OMI, which can record my screen, record my conversations, take notes on me all day, and then it combines that with something called Obsidian. Now, Obsidian is a free app that basically allows us to create markdown files and essentially document everything that I'm doing all day. Right now, I can connect that as a second brain to Claude and Claude can also help me organize it. Right? So, it works in two different ways. Number one, Claude can help me organize Obsidian, but then at the same time, Claude can pull in the context from Obsidian and the system with OMI. Right? So just to recap, OMI takes notes on what I do all day. Obsidian starts uh documenting that inside markdown files and then my AI agents can update and pull in the information from that Obsidian vault.
Right now, if you want to see what OMI looks like, this is an example of OMI, right? So you can see here it's got like all my tasks, my goals, what I'm working on, what I did today. I can actually ask it anything at any time. It's got all my memories on me and what I'm doing and it just takes notes on me as you can see, right? It also sets up tasks for me. It links to different apps. And the one that I like to link to is Obsidian, right? So, I can export all the memories and notes that Obsidian uh that OMI has taken on me. I can pull that into Obsidian. And then with obsidian we can then go into claude and use that for context. Right? So how do you do that?
Well if you actually try and pull this in inside a plug-in you'll see that obsidian is not there. Right? If we go into connectors obsidian is not there. Right? So how do we pull this in? Right. Well, what you can actually do here is you can use something called uh MCP Obsidian, right? And this will link Claude to Obsidian directly. That's one way to do it. Now, another way you can do this is you can go into a new session inside Claude like you see here. We can open our folder. So, I'm going to open up the Obsidian Bolt, which I've got stored locally. All right. So, you can see it here, right? And it's stored locally. I can open that up inside Obsidian. And then I can say, okay, can you see my Obsidian vault and notes?
Right. So, it's beginning to review that as you can see inside this chat.
So, it's going to list the contents of my Obsidian Vault, all the information, and it's got access to all my content, right, which is really cool. So, it's got like notable notes, um, when it was updated, different files, etc., and and that's basically linked to Obsidian. So, this Obsidian is the second brain, which stores all of my details. And then using Obsidian, I can then link that to Claude. Pretty simple and easy. That was very easy, right? And it has all the context on me and what I'm working on and that sort of thing, right? So I can say, okay, can you check through my memories and pull in some relevant ideas of how you could help me automate stuff dayto-day based on the memories vault inside Obsidian.
So it's automatically got that second brain context plugged into Claude. Now I would say this is far more useful than if you use Claude as the main document for your second brain. The reason for that is if you're using OMI is taking notes on your day. It understands you.
It sees what you're working on etc. Right? It's got a lot of context. It whereas for example Claude is not going to run autonomously like that taking notes on you. Right? And then once that's done, it can read my memories file. Right? Now, one of the things you're going to notice inside Obsidian is that it's a little bit messy here. A little bit messy. Right? So, we can actually use Claude to organize our second brain, which is what I'm going to do in a second. So, it's reading the memories file, pulling in the key context notes. We can also link this to any other AI agents that we use. So, for example, that could be OpenClaw. It could be, for example, Hermes agent. And bear in mind like the more information you give your AI agents, the more context they have about you, the better the outputs and the more useful they're going to become because they understand you and what you're working on, right? So with this memory MD file is it's got way more information on me than it ever had before, right?
And it can also save a few facts to memory and give me the symphysis, right?
Which is really cool.
So now it's pulled in the information from me, what I work on. It's like here's what I've pulled in from your vault and what I think you could do, right?
So, for example, it could build a 6 a.m. job that reads new OM entries and writes to obsidian with five sections.
So, it could be content ideas, decisions, distractions, wins, errors, etc. Right? It could also, for example, um take all of the SEO work I'm doing inside Claude and write that to an obsidian bowl, etc. Right? Now what we can also do here is I could use it the other way. So I can say to Claude, okay, based on my obsidian vault, can you organize it into a structured second brain? So it's much more useful plus organized. Right now it's super messy.
So for example, untitled documents not very structured.
Everything sits inside the memories document etc. Organize it beautifully.
So you can see here it's like restructuring the vault to make it much nicer. And then it's going to create a maps of content that we can build. Right? So it's a navigation process. So it's actually useful, not just tidy. Right? Um so there's two things it's doing there.
Number one, it is organizing notes and number two, it's creating a navigation map that we can go around, right?
So, it's organizing the restructure now, which is great. By the way, if you have any questions about this stuff, like feel free to ask. Um, post it in the comments if you're watching live and I'll be happy to help you.
So, you can see it's now running the new structure, setting that up, etc. And look at this. So if we go into Obsidian now, it's way more organized, right? It looks a lot nicer. So it's actually set up like daily projects, areas, resources, memories, and archive, right?
So it's it's created something that's way more useful. And then it's organized the resources into separate files and folders, right? So, for example, the AI profit boardroom assets like the offer details, the sales page.
How much nicer does that look versus like what we had before?
So, it's literally organizing our second brain for us automatically, which is great.
And now it's creating the navigation layer, right? The dashboard, the read me, the maps of content section for each section, and a daily note template.
Right. And it's it's creating like this nice graph view as well now. Right. So everything's getting organized.
Everything's becoming a lot more tidy.
We'll come back to that in a second.
That'll be something to work on.
So if we go to the home section now, it's created the second brain section, right? Restructured it. It's actually put in the date of when it's organized.
And then it's created this right now section, right? So it's got the like the inbox today and the memories. Then it's got active projects, areas that we're working on, different areas, and ongoing responsibilities, the resources. And you can literally see this being updated in real time by Claude. And then it's created like these different playbooks here, a nice archive and conventions as well. Right, it's using something called par which is projects, areas, resources, and archive. And then it's numbered everything nicely, too.
Pretty amazing. Now, if we go to the read me here, it's actually got this whole structure and it it explains like here's the rules, here's where things go, so how to organize it. So, it's great like an SP on how to organize it, which my other agents can read and understand and use in the future.
And you can see it right here, right?
Which is pretty amazing now.
If we go back over here, that's beginning to structure everything nicely, right?
But there's still some parts of it that I'm not 100% happy with. So, if we go back into obsidian over here, let's see what we got. Yeah. So, we got like for example the operation section here, all the details etc. Then it's got links to relevant topics, the strategy, um, and everything's just linked nicely together, which is great. Right now, what I can also do is you see this graph view here? It's good, but it's not really linked together properly. Right?
So, what I'm going to do is I'll go back into Claude in a second.
And I'm going to take a screenshot so that Claude understands. Okay, this is messier than I would like it. It could be a lot more organized.
And then I'm also going to take the documentation on graph view as you can see here.
I'll say documentation on how to use graph view.
We'll say what can you do here?
Right.
Let's see what questions we got in the meantime.
My man, good to see you here, Eric.
Thanks for joining.
Good morning from Switzerland. Welcome to here. Vic says, "Bro, you're a beast."
Thanks, man.
Simbo says, "I'm finally am I going to finally install Hermes today?" Let's hope so. I hope so, too. Uh, how much this cost? So, OMI is free to use as long as you don't use the chat. If you use a chat, then it it, you know, this is a subscription, but you don't need to use this for the second brain process.
And uh Obsidian is free too, right? And then you've just got Claude.
I came just in time. Been looking for this content. Happy help. Tony says, "Hi, Julian." Good to see you here, Tony. I'm curious about features. I just joined how to create to be more organized. Nice.
Eric says, "Question, sorry. Simbbo says, "Is it as effective to prompt in French as in English, in Hermes, or uh you can use, you know, there's multiple languages you can use with it. It's it's designed to be working with other languages. Um I think some more like less commonly used languages, for example, like Thai, you know, speaking Thai doesn't work so well with Hermes.
But if you're, for example, just using, you know, basic languages like French, Spanish, Italian, etc., you should be fine.
So, it's now going to start organizing the map as you can see here and it's going to color group the folders, right? So, everything's like visually more interesting and more organized.
So you can see the graph being organized in real time here.
And then it's all beginning to link together as well nicely now too.
All right. Right. So, we got the home section that links to all these and it's creating the graph.json.
Now you also see on the right hand side when you're opening up Obsidian the for each document and each note that you're creating it's got the heading and then the H2s underneath right so the subheadings which is quite nice you can sort of like flick through the document like that just look at the main parts right which is super useful So it looks at like what's working, what's improving, etc. Even like for example for our Hermes content that we were working on before. And so it's really nicely organized now. It's looking a lot better. It's even got like nice emojis for each part and each section as well, which is cool.
Can Hermes be as agentic as Google antigraity? I would actually say that Hermes is a lot more autonomous and anti-gravity.
I don't know that many people using anti-gravity for agentic tasks whereas something like Hermes agent I mean like you know the whole world's using it right like it's the most popular AI agent on um on open router.
So I would say Hermes is way more aentic.
And then you can see like the map is is getting more organized here. So that's basically how to use Claude as a second brain organizer. And then you can use OM and obsidian for actually storing content inside your second brain and also Claude Compul in that context. Hermes agent openclaw codeex etc. They can all use that similar content as well and it just makes it way easier to organize right. I think there's a lot more you could do to that. So, like for example, if we're looking at Obsidian here, one thing that I'd probably recommend is that inside, for example, each of these sections, you add more emojis, you add more like nicely organized sections, you make them link together better, etc. But it's looking better and better, right? It's gradually getting a lot cleaner, and you can see how things interlink much better now versus before, which is great.
So, thanks so much for watching. If you want me to answer your questions like I did for Richard here on video with a full breakdown and tutorial, you can get that inside the AI profit boarding link in the comments description or go to the profitboarding.com and we have a full tutorial on the infinite context engine which is the whole system that I've shown you today. So, I walk you through the whole process step by step. I'll give you a video tutorial on it. There's also 100 prompts and a 30-day road map that you can use with this sort of stuff. And that is exactly how to use Claude as your second brain. Now, the other cool thing here as well is that inside the profitable community, you can ask questions. You can connect with 3,000 members who are also growing and learning. Like you can see it's a very active community. Lots of people helping each other and posting more stuff.
Inside the map, you can connect with people in your local city who are using AI automation like Claude and AI agents.
Inside the classroom, you get access to all of my new daily training. So, we drop new daily trainings and tutorials over here. We have a full section on how to go from like beginner to expert with AI automation in just six weeks. Over there we have a full course on how to get more clients to the agency. You can learn how to rank number one inside Google and AI search engines with the AICO automation section. And then we also have like our coaching calls recording. You also get four weekly coaching calls where you can jump on live calls, get help and support in real time, share your screen, um ask questions about your setup, etc. And that's all inside here. Plus you can connect with me personally here too.
So today we're going to be looking at how to use codeex or claw code and local models with computer use. Right now if you want to run local models with codeex or claw this is actually a question from car saying like you know it's a long subject but basically what he's trying to do is figure out how to run either codeex or claw code with a local model. um but also run it with computer use right and how to use it directly here. So there's kind of two elements to this, right? Number one, if you want to run local models with claude or with codeex, you can do that via OAM, right? OAMA is the simplest, easiest way to do this, right? So the first step is that you want to download a local model.
Now, if you want something that's really powerful, you need a really good setup.
And what I mean by that is, for example, like, you know, you want a a DGX Spark or something like that, right? To run really powerful local models. and that's how you're going to get the most out of it. If you don't have a really good local setup on your computer, you know, even like a Mac Studio or a Mac Mini is okay, but it will just run Gemma 4, then you can you can use one of the cloud APIs. Right? So, that's step number one is make sure you have a good setup and then download a local model from OAMA.
Right? Now, once you've done that, um, so for example, if you wanted to download Quen 3.6 6 plus you can just make sure you have O Lama downloaded.
Once you have O Lama downloaded then you are going to copy this into your terminal. Right. So we've got O Lama over here and we just go to terminal paste that in. And that will actually pull in and download the model. I'm going to stop it um because it will take quite a while and it takes up a lot of memory. But that's basically step number one. So you download Alama, you pull in the local model as you can see right here. Once you've done that, you can then set up for example like claude code or for example codeex using these local models. Right? So for example, let's say we wanted to use a local model with something like cloud code, right?
We could I'm I'm going to do it with a cloud model, but it's the same process with a local model. So let's say for example I wanted to run it with DeepSeek, right? Well, I just copy and paste a command like this and it pulls in claw code, but it's running with DeepSeek, right? That's the biggest difference. We can do the same thing with codeex. So, if we paste this in, we can now run deepseek with codeex, right? And so, you can run these local models with codeex directly there.
Right? That's how you can use these models. Now, if you want to use computer use, you would just give it the details on exactly how to use that, right? So there's quite a few different because bear in mind like you're running this inside the terminal right if you on the claw desktop it's a totally different scenario but if you're running this inside for example um claude code with a local model then you would give it the instructions to use computer use so I'm just going to check can you use computer use inside claw code with deepse v4 And we'll try the same thing inside codeex as well.
All right. Now what we need to do from here is we need to use something called CUA driver, right? which is the skill that can open and use computer use right so for example here I'm going to say use cua driver skill to open up Google notes uh sorry to open up notes application right so you can see I don't have this opened yet but we're going to use cua driver which is a skill for computer use with this model to open this Then we just need to allow the settings with CUA driver which I'm doing here.
All right. So we've just allowed that And now it's open up notes, right? As you can see, Let me try another example here. So, if we go inside here and we're like, okay, write a note about today inside notes, right? And I'm just going to close notes again.
And we'll say using CUA driver.
And the cool thing about using computer use with CUA driver as a skill is that it's running in the background and it's working without me, right? So, I can carry on working on Claude inside the tunnel. I can continue doing whatever I'm doing without getting interrupted by the computer use. Right?
So, it's pretty powerful here.
Now, honestly, is like computers amazing on on any sort of platform? I don't think so. Um but it is quite it is useful to run in the background and do other stuff you know like it can get stuff done you know Hermes can do the same thing. So if we go inside Hermes here using CUA and we say okay use computer use and write a blog about Hermes inside notes application right it's going to prepare computer use as you can And now it's opening up notes. So this is notes over here.
And you can see it's navigating on the screen and starting to write on what it did. Right? It says it's May the 14th started working with CI driver today to automate Mac OS apps from the terminal.
Right? Notes can be created and notated blah blah blah. Here's the key takeaways. Right? And so you can see here that it it created a full note using OAMA with uh DeepS and CUA driver.
Right? So it can it can create notes. Is it smooth? Is it easy to use? No. But if you're insistent on using computer use, that's how you can do it and that's how it works step by step, right? Pretty cool. So just to recap on the steps there, you make sure you have a good local setup. Then you're going to download O Lama. Then you're going to download uh local model.
Then you're going to run claw code using lama set up cua driver and use cua driver for computer use. That's basically it. That's the whole process.
So, if you want to get a full training on how to use computer use, we actually have a full section on using it with Hermes agent which we created yesterday.
You can see that on Hermes computer use here. It's a full video tutorial and step-by-step guide as you can see and that shows you exactly how to use computer use step by step using these AI agents. And if you want me to ask answer questions like we did today for KL on a video, then feel free to join the powerful board. You can post inside the community. I answer questions like this every day. Um, you can ask questions, get help and support whenever you need to. You can get new trainings inside the classroom on all my new stuff. You can see all my new daily updates on the left hand side here. We also do four weekly coaching calls where you can jump on live calls, share your screen, ask questions about your setup, etc. And then inside the map, you can connect with people in your local city who are using AI agents like you. So feel free to check that out. Thanks for watching. I'll see you on the next one.
Cheers. Bye-bye.
Vidéos Similaires
BREAKING: Microsoft’s New Image Generating Model Beat Out GPT 1.5 and Nano Banana 2
aimmediahouse
122 views•2026-06-03
Long-Running Agents — Build an Agent That Never Forgets with Google ADK
suryakunju
142 views•2026-05-30
I Made the Same Anime Fight Scene in Every AI Video Generator
NobleGooseAnime
295 views•2026-05-30
Nvidia Bets Big On AI PCs | New Chip To Power Windows Laptops | Technology | AI Updates | N18S
cnnnews18
3K views•2026-06-01
3D Platformer Update - NO CAPES
SolarLune
294 views•2026-05-30
AI Doesn't Create Bias — It Inherits It
UXEvolved
176 views•2026-06-01
Distributed Inference Challenges Explained #shorts
alexa_griffith
466 views•2026-05-31
[한글자막] OpenAI @ Replay 2026 | OpenAI는 Codex로 개발 방식을 어떻게 바꾸고 있을까요?
TechBridge-KR
1K views•2026-06-03
