Codex + Claude Code + Agent OS Is INSANE!

So today we're going to run through the latest questions I've had on the agent operating system which you can see we've built out here. This is a powerful system you can use to basically build and automate anything, link all your agents together. You can use it for example for lead generation and outreach campaigns. You can voice control it and voice activate it with Hermes Jarvis.

We've even got a memory galaxy so that all of our agents here know exactly what we're working on, what we're building. We've got Kanban boards where we can orchestrate agents. We have for example a local AI agent engine here where we can build and automate anything that we want locally as well. And so today what I'm going to show you is some of the latest questions we've had on the agent operating system and then answer them because I know if people have these questions inside the AI profit boardroom then you probably have these questions about building agent operating systems yourself.

So let's get straight into it. So first question we have was about codecs. So Rob was saying you know brand new here just signed up watching videos but can you use with the agent OS your codec subscription and how can you use it right and can you use it with codecs too. So we do have a codec section over here.

So we have the chat and then you've got the goal mode and the workspace and all your previous sessions saved as well so it's easy to see everything you've built. So that's one way of using it and then you've got Hermes as well. Now with Hermes you can use GPT 5.5 with codecs login and you can use that for generating images as well. So that way you can use the power of GPT 5.5.

You also get access to the studio where you can generate images videos as well using GPT 5.5 and then you can also plug it into for example our open core or anything you want. As well as that actually what happened recently this is a new update that just dropped out a couple of days ago is there's a plugin for codecs inside Claude Code which means you can actually use codecs inside Claude Code to power up too. So there's like four different ways you can use it here. So just to recap yes you can and welcome here.

So you can use codecs in multiple ways inside the agent OS. Method number one is you can plug into Hermes agent. Number two is you can generate images with OpenClaude and Hermes agent 2 using your codecs auth. You can also use codecs inside the codecs tab of the agent OS and additionally Claude just dropped a plugin where you can use codecs inside Claude Code as well.

So you have a good post here from North as well who's talking about how it's better to use systems rather than focus on prompts and this is 100% it right. So he was saying basically the real skill is building repeatable systems and improving what's already tested and having the patience to debug them properly when a new update arrives. This is 100% it. So for example like GPT 5.6 just dropped but it's only available to certain people but the good thing is it doesn't even impact any of our workflows anyway because we can swap in the model swap them out.

So for example as well when Sakana Fugu dropped earlier this week we could plug it straight into our systems and build some amazing stuff with it and we could only do that because we focus on systems not models. So I think that's the long term play here. Rowser says the best. Julian you have one of the best training and one of the most accessible AI experts out there which is pretty nice.

Number one thank you so much and I think number two like the whole point of building this community is to share and learn and grow together so the only reason that I've developed these amazing systems for example like the AgentOS is because I have a great community of people who encourage me and you know everyone who watches stuff like this they help and inspire me so the only reason that I've achieved what I have today which is probably quite little in itself but you know we've got a great community and a fantastic group of people the only reason that I'm here is because of all the people that help me inside the community. It's a good question so the question is like Quen 3.6 versus Quen Agent World has anyone tested it etc and I was looking at this announcement yesterday for Quen Agent World the interesting thing is I can see a lot of people talking about it but I can't see anyone building with it that was pretty funny when I saw the announcement I was like this sounds amazing but what if people actually built with it so from what I've seen I haven't seen anyone like build anything interesting with Agent World yet so it kind of seems more like theory than people actually building with it directly but I know that Quen 3.6 particularly is a really good local model if you have a great setup. Nacho was asking about how to connect Codex and Cloud Code with a local LLM and how to do that so there's two ways I know I know the Codex basically have their own documentation on how to connect local models as you can see right here so you can change your model like so and you can change the local environment from the config.toml file and you can change your model around like that and then another option that you can have is you can use Olama now GLM 5.2 on Olama is cloud based but you can basically run this in one click with Cloud Code or for example Codex as well so there's two options there. Davide was talking about how GLM 5.2 is a genius and I found the same thing I mean he talks about a use case that he's used to do this but I found like GLM 5.2 is is really good I mean like it's it's definitely up there very comparable to something like Opus 4.8 and we actually ranked all the models and GLM 5.2 was right up there in terms of how it performs so you can see for example on GoldieBench like I personally test it and I show everything that I test it and host it so you can try it yourself but some of the stuff we built this was absolutely amazing like and when we tested it side by side versus other models for example like uh Claude Opus 4.8 it was comparable like it was it was actually really really good so for example if we go to the compare section on GoldieBench you can actually compare the models side by side to see how they perform and GLM 5.2 was like if we have a look here like compared to something like Kimi K217 is miles ahead so this is for example the output from GLM 5.2 this is the output from Kimi K217 and obviously Kimi K217 wasn't anywhere near as good so from all of my tests so far it's right up there and it's actually outperforming on some of my tests versus something like Opus 4.8 and also something to bear in mind is like if you look at the release of GPT 5.6 which is gated according to rumors and also if you look at the release of Fable 5 they got taken down like it's amazing that GLM 5.2 is open and they've published open weights for it whereas for example if you're using GPT 5.6 well you can't get access to it or Fable 5 you could get access to it for three days but now you can't again and we don't know when it's back out so it's interesting to see these open models overtaking frontier models as well this is a good one so Ben is talking about the AgentOS he's like you know thanks for the support your responses alone have saved me so much time on what we've done so far but he's at a crossroads with his Hermes setup so he said he's built so much that he can use it but he's failing to give it to his team so that they have something they can use right so there's a few ways he could organize this so for example he's talking about the AgentOS and like how he could get give that to his team but he doesn't know how to simplify it or to make sure that they don't get distracted or hit bugs etc so this is a really good question what I would recommend is that you could create a simplified version if you don't want them to be distracted so for example you could have the AgentOS but you simplify it down to like the workflows your team actually needs and use and then you quality control their work just to make sure they're only using it in the right way so for example like you don't need to give them access to all the CLIs you don't need to give them access to all the workflows inside the self section you could just have like one CLI for example Hermes and then you could have a couple of workflows that each of your team members focuses on or you say okay only use this section inside my AgentOS and ignore FNL the other option is that you could write an SOP for them so that when they use the AgentOS they just follow those instructions and rinse and repeat those step by step that way they don't look or even get distracted by everything they just focus on the step-by-step commands you've given them and you could just write that up in a google doc could be like five steps and then from there they only use that part of the AgentOS and everything is simple for them when you actually set up the AgentOS for your team as well what you can do with the version you give to your team is with the agent you've used to set it up you can say okay remove this or remove that etc there's actually a settings section in the top left where you can hide certain parts of it and then save it so for example we can hide open claw then click the tick button once we've changed the settings and that will hide it and simplify the UI as well so we've got a good question from Stefan here and Stefan was asking like you know how to make videos to promote his SaaS tool and so he has everything recorded no problem doing the voiceover but he's like stuck on the what's the quickest way to put these together phase right so for example if it's a quick like 30 second to do how to to do section so there's a couple options you could do here number one you could use something like the video agent inside the AgentOS that's an option if you don't want to set that all up and I'll just show you an example here so you can see for example if we look at this video this is fully put together with AI with the avatar with the b-roll with the script the research etc that will automate it inside the AgentOS if you've got some footage you just need to put it together you can use a skill like hyper frames and Hermes agent as well or Claude as well so if you just want to for example edit a certain part of it you can get Hermes or Claude with the Reemotion skill which is free and open source and then you can train your agents to actually edit the videos for you and put them together we actually have a full training right here on how to do that and the cool thing is you can give the agent the footage that you have and then just ask it to edit the certain parts for you and put it together so this is a good question you know what would we use for each setting so basically this member is asking you know I'm curious what cloud models you're currently using for which specific tasks so mainly interested in setups that stay within the OLAMA ecosystem whenever possible so I think if you were only using OLAMA obviously that limits your options like by quite a lot I would just focus on probably GLM 5.2 right if I'm looking for all the models and I want to simplify I'm just going with GLM 5.2 because it's the closest you have to frontier level inside the ecosystem of OLAMA.

Minimax M3 is good but it's older, Kimi K Super 7 is still good but it's not as good as GLM 5.2 so if you have a choice out of all these just go with GLM 5.2 if you want to stick with OLAMA and I think that's probably the best model you can opt with then you can plug it into Hermes, into OpenClaude, into Claude, into FreeCode, whatever you want and also the good thing about this is that you get used to using that one single model, you master it and that way you get better outputs every time you use it. So we've got a question from Shane here and he's basically asking like how do you keep the AgentOS updated right so we share our AgentOS system with all members inside the AirProfit volume but how do you keep it updated how do you get the latest versions with the new features and everything else so if you've got an older version you need to update it it's inside the zip file or the new version so you can see the last update is 26th of June we've got the resources right here and basically with that you can easily get access to this so you can use the update md file inside the zip whenever you want to install new features bear in mind like you don't need to install all of them you can just pick and choose the ones you like. Davide is asking you know there's a skill called remember with Claude do I use it personally I actually skip this so what I do is I just use my obsidian vault with the memory galaxy and then I plug that into all my agents and then for example if I ask Claude something like you know what did we recently work on or what did we work on last week it can recall the dates and it recalls a lot of data because it looks for our obsidian vault so that's what I use instead to make sure that we give Claude a really good memory and we have a full system on that inside the infinite context engine module that shows you exactly how to do it. So another question here which is Hermes profiles for the team right and basically Seraph is running into multiple issues whilst creating profiles in Hermes and it prompted him to create multiple gateway instances and he's creating a Hermes installation of his whole team and his goal is to have a shared brain for the company which all the team has access to so for setting up agent profiles we have a full guide on it as well which I'll send to you but also what could be pretty interesting for the shared brain you could set up obsidian and then just have a shared obsidian vault across your whole team and that way they can all access the same notes the same memory the same system but they can update and improve it for you as well and that way you have a shared knowledge vault across your team if you want your team to have access to it remotely you can actually create a login with obsidian share it with your team and they can all share the memory together another option is that you can connect it all together with an MCP so that your Hermes agent has access to the obsidian MCP but I think just having a login in the cloud for obsidian would be a much simpler option and we've got a full training on that here and that's basically every question answered boom shakalaka all right so we have answered all the questions on the agent os to get today if you want to be able to ask me questions like this and I answer them for you in daily video tutorials like you've seen today I answer all the questions personally inside the AI profit boardroom so you can post a question there and then I'm happy to help inside the classroom get access to all of my new trainings as you can see and we have the agent operating system with a video tutorial last update date you can get the zip file as well to install it and that's all inside the classroom here as you can see and then we add new daily tutorials based on what's just dropped that's actually useful for you and then inside the calendar you can jump on weekly coaching calls get help and support whenever you need to and inside the map you can meet people in your local area from the community who are building with AI agents and building the agent os like you can see today you also might be thinking okay this stuff sounds technical is it difficult is it easy to install etc so I've set it up and I'm non-technical Rick set up his agent operating system from the AI profit boardroom in like 30 minutes you can see for example browser is absolutely loving the community and Matteo got his agent operating system set up pretty quickly so I know that if so many people can set up their dashboard and I can too then everyone can win with this and you don't need to be technical to get this set up so hope to see you inside there link in the comments description or just go to the aarprofitboard.com thanks for watching bye