← Back to search

I Built A Real JARVIS With Claude

AI News Today | Julian Goldie Podcast · 2026-06-24 · 13 min
relevance 67 2747 words Episode page ↗ Audio ↗
Show full episode description
Hermes Agent OS Q&A: Local vs VPS, Obsidian Memory Vault, and Automating Content in One PromptThe speaker showcases an “agent operating system” built around Hermes, including a real-time voice-activated Jarvis mode, Hermes Oracle for web/news research and content drafting with source links, and a connected “memory galaxy” second brain powered by an Obsidian vault. They explain they run the system locally (not on a VPS) for security, while relying mostly on cloud models via APIs/CLIs/OAuth (e.g., ChatGPT Realtime API, Groq OAuth, Claude via CLI), noting local models typically require high-end RTX GPUs. They describe hosting files locally and in the cloud while keeping the database (Obsidian) stored locally and auto-updated by agents. The episode answers viewer questions, demonstrates automated SEO/article deployment and social content creation from trending news, emphasizes focusing on systems over specific models, and invites viewers to get the ready-made Agent OS inside the AI Profit Boardroom with tutorials, community help, and coaching calls.
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
Viewers want to know how to safely build and run a JARVIS-style Agent OS; this Q&A covers hosting, models, and mindset.
Benefits
  • Voice-activated real-time Hermes Jarvis assistant
  • Local sandbox setup avoids VPS security risks
  • Auto-organized Obsidian memory vault as second brain
  • One-click content and AI avatar video generation
  • No coding skill required to get started
Use cases
  • Hermes Jarvis voice chat via ChatGPT real-time API on Hermes
  • Hermes Oracle uses Grok OAuth to pull news and draft content
  • One keyword and case study deployed articles to 5 websites automatically
  • Rick set up the Agent OS in 30 minutes; Jose in under a day
  • Built outputs with Sakana Fugu, tested on Goldie Bench
KPIs / results
  • 5 sites published from one keyword
  • 187 pages of member testimonials
  • Setup in 30 minutes to under a day
  • RTX 5090 minimum for local models
Tools / build
0:00 / 0:00
📑 Chapters — tap a time to jump there
00:00
Agent OS Overview
  • Agent OS with real-time voice Hermes Jarvis
  • Hermes Oracle, memory galaxy, video agent built in
01:13
Local vs VPS Setup
  • Runs locally on Mac Studio, not VPS
  • VPS scraping caused OpenClaw problems; models via API/CLI/OAuth
04:40
Automated Content Engine
  • One-click content deployed to 5 websites
  • Hermes Oracle drafts social posts from news
06:27
Skill Level and Mindset
  • No coding needed, everyone learns from beginner
  • Focus on systems, not which model wins
So we have built out an agent operating system with Hermes. It is crazy powerful. We have, for example, this real-time version of Hermes JARVIS where we can chat with it. We can put it in warm mode. It's voice activated. It works in real time. We can speak to it and also has previews down here. Now, we also, for example, have Hermes Oracle. Hermes Oracle basically can consult the web, find the latest news information each day, link to the actual Twitter post that it's related to, draft content, and find me an idea for that content as well. And then on top of that, of course, we have Claude built in. We have Hermes built in. We have a memory galaxy where we have all of our memories connected inside one powerful context system as a second brain. And we can basically build and automate anything to the point where, for example, we've got a video agent right there that can generate AI avatar videos on the spot, edit them, script them, research them, write them, and create beautiful B-roll for them in one single prompt. That is a game that we are playing right now with the agent operating system. And it's mind-blowing stuff. And this is all inside the AIPROF reporting. But what I want to do today is answer some of the latest questions I've had recently about all of these systems that I can help you as much as I can. Because I know, like, if people here have questions like this, you probably have questions as well. So let's just get straight into it and start answering all of the latest questions that we've had about the agent operating system, how we build it, how we use it, etc. So the first question we have here is, are you running this all locally on your computer and giving full access to your computer? Or is it a VPS that you run the agent operating system on that you can just access from a local host dashboard? Pretty impressive stuff. Well, number one, thank you, Jimmy, for that. Number two, when it comes to the setup, I run it locally. I like to have a sandbox just on a Mac Studio so that, you know, I don't think it's very good to have that connected to the whole internet. And I think that, you know, if you look at all the problems that OpenClaw had, all of OpenClaw's problems came from people on VPSs. Because, like, people are actively scraping VPSs and, you know, sometimes getting login details and stuff like that. So you need to be super careful about how you connect this stuff. So I have it running locally. I don't use a VPS. You can use a VPS. There's actually a training inside the AI Profit Boardroom on how to use a VPS if you prefer that. But for me personally, yes. Just run it locally. Having said that, the models that we run with this are mostly APIs and CLIs and OAuths. So, for example, we look at Hermes Oracle. That is set up with Grok OAuth on Hermes. If we look, for example, at Hermes Jarvis, that is basically 11... No, sorry. ChatGPT real-time API. And it's plugged into the Hermes agent that we have running. And so I've run all the models with cloud models mostly. I just don't think local is that good unless you've got an RTX or something like that. Another question we've got here. Where do you host all of this stuff inside the agent OS, right? So, you know, when we're looking, for example, at the agent OS, where do you host it? Well, I have a local setup for hosting. Then we use cloud files. And then for actually storing the database, which I know you've asked about here, blank. Good name. For storing the database, we have that all stored locally. So we have an Obsidian Vault. I've never touched any of this, never edited any of it. My agents just run this and organize it automatically for me. So it looks beautiful. And then it's plugged in here. And if we actually have a look at our Obsidian database, this is a free app you can run locally. You can install it into your agent operating system. That means your agents have context on you. And they can also update the Obsidian Vault every time you use it. So, for example, if we look at the vault here, these are our latest memories. You can see, for example, me and Hermes Jarvis speaking. This is a full transcript from the conversation between me and Hermes Jarvis. And so when you're building this out, the whole database here is stored with Obsidian. And I store it locally, but we have a login so that we can log into our account from anywhere as Obsidian and then restore the details. So we've answered Jimmy and we've answered blank. Let's have a look at the next one. It's called 25GB dual RTX 590. Yeah, or 9090. Well, yeah, so you could run that with local models. That's probably the best way to get the most out of local models. From what I've seen with members inside the AI Profit Boardroom, like the guys who are creating, like, for example, avatar videos and AI avatar videos and AI images locally, they're running RTX 5090s minimum. So, yeah, if you have that setup, you can run local models. But I don't run local models. I have a Mac Studio. I use free APIs. I use OAuth. I use CLIs. Stuff already got, you know, running in the background that I've already paid for. So, for example, like, Claude is just plugged in here with CLI. Like, you don't need to use API or anything like that. It's pretty cool. What else have we got here? So, Rav, I'll be following along. I am hopefully building a space weather for heliobiology news plate service starting from social media posts and turning into mind-blowing published articles and magazines for live updates. And I think I might just use Hermes to give me the tools to start making it easier to build it. That's pretty cool. That sounds like an interesting project. Now, one thing that we do is we actually have Hermes posting content for us inside the agent operating system. So, if we go over here to the SEO tab, you can see that we can research from our existing Google search console. So, I know you're talking about using a website and building automatic social media posts as well as websites. So, we can basically generate content in one click with a keyword and a case study. And what that would do is deploy it to our website. So, if we look at the history of deployment here, we created this article and five others to five different websites automatically with one keyword and one case study. Beautifully formatted. I can't log into this website, but this whole website is fully automated with AI, right? And so, it's unique to me. It's not like spam or anything. It's actually got useful case studies and it's actually interesting. But the point to note is like, yes, you can do that. And it's the same with, for example, social media content. So, we have a video generator over there that we can use to generate that. And we also have Hermes Oracle where you can look at the latest news, as you can see right here. We can open it up and then we can actually quote the news and create some content around it. So, you can see here it says draft this. And we can look at the latest news, quote that content, create social media content around it automatically in one button. And then we don't need to find what's trending. We don't need to find new topic ideas. We don't even need to think about what type of content we're going to create. It's all automated inside this beautiful system with Hermes Oracle. So, that's pretty cool as well. Now, this is an interesting one. So, Control Alt Pete says, I think I am at AI development level five, which I'm not even sure what that is, to be honest with you. I mean, I think we're all just learning. This technology is so new that nobody fully, truly understands the power of it. And I say that as somebody who's built incredible systems around it, like you can see with Hermes Oracle right here and the whole agent operating system. I'm not a coder. And even coders that I know personally feel like they're behind with AI. So, if you feel like, oh, getting value out of something like the agent operating system is too technical for you, I would say that you are missing out. Because, you know, the sooner you start cracking into this, the sooner you'll start learning it. Everyone learns from beginner because it's so new. There's no courses on this. There's no, I mean, there's my course, but, but I mean, like, what I mean is like, there's no courses at university. There's no formal education for this. Everyone's learning together. This is brand new. Everything changes day to day, right? And so, you don't need to be at a certain level. And I'll give you an example of this. I'll give you several examples of this. So, if we look at some testimonials from the AI Profit Boardroom here of people who have set up our agent operating system, and we've got 187 pages to this stuff, right? Rick just joined the school and he set it up within 30 minutes. I'm not technical. I created the whole thing. And you can see, for example, Jose set it up in one single day, like less than a day. Loads of people are creating these AI agent operating systems and dashboards, and they've never used AI or never been technical before. So, my point is that if you can click a few buttons and if you can use Claude, that's all you need. And you'll learn from there. And we all learn and we grow from the same point because this is so new. No other generation in human history has ever had access to this sort of technology, right? And so, this is brand new to us all. And we can all learn. We don't need to be technical. Just take one step at a time. Enjoy it. Enjoy the journey. You're early. This is interesting. Opus 4.8 front-end design and good prompting blows Fugu from Sakana. AI slot website away. I don't... Sometimes I see these comments and I'm like, I can't... Number one, I don't understand them. But number two, it's like, I don't look at which model is the best or which model you should be using. I look at what is the best system based on what comes out. So, for example, Sakana and Fugu just got dropped yesterday. Fantastic AI. Created some absolutely insane stuff. We're already testing it on Goldie Bench. And the benchmarks are insane for this stuff, right? And if we show you some examples of what we built with Sakana inside our workspace here, some of the stuff that we created here is just absolutely outrageous. It's really, really interesting powerful stuff. This was just something we created yesterday for fun. You can see another example right here. And it's absolutely wild. Now, I don't care which models are best. I don't really care what comes out and what comes in. What I care about is how do I automate my work and get the best possible output? So, that's the way that I look at it. It's like, I wouldn't even care about the models. I'd care about the systems you build. And that's why we built the Agent OS system to get the most out of stuff like this. And, you know, I think for anyone watching, if you're wondering, okay, like, should you be using Sakana Fugu? Or should you be using Fusion? Or should you be using GLM 5.2? Or should you just wait until Fable 5 comes out? I would say don't worry about any of that. Build the systems to automate your work. Just do that. And focus on that every week. That's what we did with the video automation here. That's what we did with the music section, the loop engineering, the SEO section here, the thumbnail generator, notebook LM that can generate podcasts and infographics in one single click. Everything that we've built here, including the memory, is set up to be a new system that makes all the tools I use, all the models. And whatever comes out tomorrow and whatever gets taken down tomorrow, whatever happens, we are anti-fragile and we can bend and be flexible to whatever comes out next. And I think that's the best way to look at it. It's like, don't even worry about the models. Focus on the systems. And I've always thought about that. Like, even, for example, when I'm hiring people, it's like, to be honest, I try and make my system so simple in my business, my agency, in the app offer boarding, that it doesn't matter who the people are that are operating it. We McDonaldize it to the point that anyone can operate it. Anyone can get the most out of this stuff. And I think that's the best way that I've seen to really win with this stuff. And just final one. Mark Phillips says, thank you for the real Julian. I'm not sure, but I no longer watch the AR ones, even though they're probably good. We don't know. I don't think that's weird. I think, like, if you just like human content, that is totally fine. And, you know, some people see my other channel and they see the AI content that we generate. And some people love it. Some people hate it. But that's why we have this daily human-only channel right here so that you get daily tutorials from me. I try and connect with you as much as I can. I do create human videos daily. And, you know, whether it's AI, whether it's human, whatever you prefer, no judgment. But I'm always going to be here and I'm always going to be creating human content as well. It's not just like because we can automate everything we do. I think there's value in creating human content as well because you can connect with people in a way that you probably can't with AI content. Right. And there's still that authenticity that people crave. So that is it for me. Now, if you want to get the whole system, I've answered all the comments from yesterday. I'll keep doing videos like this if it does well. If it doesn't do well, then I'll just be like, you know, maybe we stop because people don't want to see that type of content. If you want to get my agent operating system, you can get that inside the AI Profit Boardroom. It comes ready made straight out of the box. You can get it from this section right here inside the new daily updates. So you get a video tutorial. You can see the last update there. You can get a zip file for installing it. And then we have daily new tutorials to help you get the most out of this stuff. So whatever drops. So, for example, like we created a new Hermes email agent today. We got a full tutorial on how to set up. We looked at Sakana versus Fable 5 versus Fusion. These are brand new technologies that are basically alternatives to Fable 5. And we've already tested it out, put it on Goldie Bench and created a full video tutorial. I answer the questions inside the community personally every single day with a video tutorial. And you can also get help from the community. There's always people online. Inside the calendar, you can jump on with your coaching calls. Inside the AI Profit Boardroom map, you can actually connect with people locally near you who are building awesome stuff like you've seen today. And I hope to connect with you inside the AI Profit Boardroom. See you on the next one. Cheers. Bye-bye.