Big Idea

The argument in one line.

Connecting Hermes AI agent to Claude Code through shared memory, Pantheon personas, and API integrations like Apollo and Gmail creates a unified intelligence system that maintains full context across both coding and conversational work.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You're a solo founder or operator running a technical AI stack locally and want to connect Hermes, Claude Code, and business tools into one unified memory system.
A developer or technical founder who uses Telegram for brain dumps and Claude Code for building, and wants context from one system to automatically inform the other.
You're already using Hermes or Claude Code individually and want to layer on lead scraping, email automation, and persistent memory without rebuilding your workflow.

SKIP IF…

You're not comfortable with local AI setup, GitHub, or connecting APIs — this requires hands-on technical execution, not just conceptual understanding.
You need a no-code solution or are looking for AI agent setup that doesn't involve running Hermes locally or wiring custom integrations to your own systems.

TL;DR

The full version, fast.

This walkthrough shows how to wire an open-source AI agent on Telegram into the same shared memory that Claude Code already builds on your machine, so neither side loses context when you switch from coding at the desk to dictating from your phone. The method is a three-step install: connect the agent via a single terminal command using your existing model credentials, mirror its config and persona files to a private GitHub repo so they survive machine swaps and can be rolled back, then bridge it to your Obsidian vault and Claude OS dashboard for cross-tool recall. From there you layer skills like Apollo prospecting and Zapier-to-Gmail, scoping each integration to draft-only and read-only permissions so the agent can act without going rogue.

Members feature

Chat with this breakdown.

Modern Creator members can chat with any breakdown — ask for the hook, quote a framework, find the exact transcript moment. Unlocks at T2: refer 3 friends + add your own API key.

Create a free account →

Chapters

Where the time goes.

00:00 – 01:30

01 · Hook + problem frame

10X claim plus credibility. Introduces the Two AIs Zero handshake pain point.

01:30 – 07:42

02 · Visual intelligence layer

Claude Code OS dashboard tour: AI spend across models, overnight dreams feature, independent memory systems, Pantheon preview.

07:42 – 11:45

03 · Installing Hermes from scratch

Terminal install command, Telegram bot creation, bot token and allowed user IDs, first confirmed response.

11:45 – 16:48

04 · Building and backing up the Pantheon

Named personas (Labyrinth/Mercury/Philosopher), GitHub mirror for config persistence, cron backup at 11PM, folder-based auto-discovery architecture.

16:48 – 19:49

05 · Obsidian agentic memory

Vault path lookup via Claude Code, memory confirmation in Hermes, real-time memory save demo.

19:49 – 21:33

06 · Claude OS bridge + morning cron

Feeding Hermes access to Claude chat logs, 8AM morning brief cron with GPT 5.5, automated overnight reflection.

21:33 – 27:30

07 · Apollo lead scraping skill

Why Apollo (B2B database, intent signals, sequences built-in), API key setup, live demo: 20 Austin roofing companies with pain-point brief and prioritized outreach list.

27:30 – 31:08

08 · Zapier MCP — Gmail + Calendar

Principle of least access applied to tool selection (draft only, no send), 7 Gmail tools plus calendar, live calendar query demo. CTA to next video.

Atomic Insights

Lines worth screenshotting.

The biggest problem with any AI assistant is that what you do in Claude Code never reaches your mobile agent, and what you tell your Telegram bot never reaches your coding session.
Bridging a local AI coding environment and a mobile AI assistant through a shared memory folder solves the context fragmentation problem that every multi-tool AI user hits.
Named personas with specific models and system prompts — a research persona using DeepSeek, a background automation persona using a cheap model — let you right-size compute to the actual task.
Never give an AI agent the ability to send emails, only draft them — the principle of least access applied to email is non-negotiable until you trust the model's judgment completely.
Backing up your Hermes configuration to a private GitHub repo means your personas and settings survive machine swaps and can be rolled back if something breaks.
Apollo's B2B database costs money for a reason: scraped data typically hits only a fraction of the contacts, while reliable paid databases dramatically improve email delivery rates.
Zapier's MCP integration turns any popular SaaS into an AI-connectable tool in under five minutes — the Willy Wonka of automation connectors.
A cron job that queries all your AI chat logs overnight and generates 2-3 improvement suggestions each morning is a free coaching layer running on cheap models.
The soul.md file is where you brain-dump everything about yourself so your agent knows your context, preferences, and goals without requiring repetition.
Connecting Obsidian to a Telegram AI agent gives that agent access to your personal knowledge base during any mobile conversation, anywhere.

Takeaway

The handshake is the product.

Creator OS playbook

The next content category is not how to use AI — it's how to make your AIs talk to each other.

Jack's Two AIs. Zero handshake. slide is the whole pitch in four words. Build that sentence for whatever fragmentation your audience feels.
The Pantheon folder-based architecture (each persona = one YAML file) is a dead-simple extensibility pattern — steal it for any multi-model routing system.
Morning brief cron is a product feature, not a tutorial trick. An AI that thinks overnight and brings you 2-3 ideas at 8AM is a different category than a chatbot.
The principle of least access is a trust-builder. One sentence in a permissions flow is the difference between this feels sketchy and this person knows what they are doing.
Jack's custom marketing page inside the tutorial — the mythology artwork, the bold headlines — elevates a screen-share tutorial into a product reveal moment. Worth replicating in your own demos.

Glossary

Terms worth knowing.

Hermes: An open-source AI agent assistant that runs on Telegram and is designed to think, dream, and self-improve based on ongoing conversations and memory.
Claude Code: Anthropic's command-line coding tool that lets developers work with the Claude model directly inside their terminal to write, edit, and ship code.
AI operating system: A dashboard layer that unifies models, chat logs, memory, and integrations from multiple AI tools so a single agent can read and act on everything in one place.
Handshake (between AI systems): A shared connection that lets two AI tools exchange context and memory, so work done in one is visible and usable to the other.
Pantheon: A collection of named sub-agent personas inside Hermes, each with its own system prompt, role, and assigned model for handling specific kinds of tasks.
Persona (AI): A named character given a description, system prompt, and model so it behaves consistently for a defined job like research, planning, or tool calling.
System prompt: The hidden instruction that defines an AI's role, tone, and rules before any user message, shaping how it responds throughout the session.
Labyrinth: A Pantheon persona built for deep research and wrestling with ambiguous problems, typically assigned a heavier reasoning model.
Mercury: A Pantheon persona used for autopilot and scheduled background work, usually paired with a cheap or fast model to handle routine tasks.
Philosopher (persona): A Pantheon persona prompted to surface the question behind the question and reason from first principles rather than answer directly.
Oracle (persona): A Pantheon persona configured for predictive or interpretive responses, defined by name, description, prompt, and chosen model.
Codex (OpenAI): OpenAI's coding agent CLI included with a ChatGPT subscription, used here as one of the model backends Hermes can route requests through.
Antigravity: An AI-assisted coding environment with a built-in terminal, used in the walkthrough to run installation scripts and talk to the Hermes setup.
DeepSeek: A low-cost open-weights language model often assigned to background or research personas where heavy reasoning is wanted but budget matters.
Telegram bot: An automated chat account on Telegram, created through the BotFather account, that exposes an API token so external software can send and receive messages.
BotFather: Telegram's official bot for creating and managing other bots; you message it to register a new bot and receive an authentication token.
Bot token: A secret string issued by BotFather that authenticates an application's requests to the Telegram API on behalf of a specific bot.
Allowed user IDs: A whitelist of Telegram account IDs permitted to message a private bot, so only the owner can talk to it and trigger actions.
VPS: A virtual private server — a rented always-on machine in the cloud used to host services that need to stay online when a personal computer is off.
Compression threshold: A setting that controls how full an agent's context window must get before the system summarizes older messages; higher values compress less often but cost more memory.

Resources Mentioned

Things they pointed at.

00:00productHermes AI agent ↗

03:22productClaude Code OS dashboard

17:50toolObsidian ↗

22:30toolApollo.io ↗

27:30toolZapier MCP Server ↗

30:30toolApify ↗

Quotables

Lines you could clip.

02:42

“There's a Hermes on the walk, you open Claude code, you type it again, it's never seen again.”

Pain point crystallized in one sentence. No setup needed.→ TikTok hook↗ Tweet quote

14:01

“We don't need Albert Einstein to wash off to mop our floors. We want Albert Einstein scribbling on a whiteboard in a very esoteric area.”

Memorable analogy for model-to-task matching. Standalone punchline.→ IG reel cold open↗ Tweet quote

22:50

“The delta between idea and average should be small.”

Quotable philosophy. Applies broadly to any friction-reduction product.→ newsletter pull-quote↗ Tweet quote

28:20

“The principle of least access — any tool connection has the fewest possible tools available to actually do the thing.”

Clean named principle. Rare security hygiene moment in an AI tutorial.→ IG reel cold open↗ Tweet quote

The Script

Word for word.

metaphoranalogystory

00:00Hermes is the world's most powerful AI agent assistant. But if you combine it with Claude code, you can create a powerful twenty four seven AI employee, solving Hermes and OpenClaw's largest problem.

00:14And in this video, I'll show you exactly how to set Hermes up from scratch and how to give it a visual intelligence layer, letting it improve itself based on how you use Claude code and vice versa. Meaning, you have a universal AI intelligence system that will save you time, make you more money, and will get you light years ahead of your competitors.

00:35And if you're new, my name is Jack. I built and sold my life tech startup with a gazillion customers. Now I'm building my own AI companies, and I just share here the stuff that actually works.

00:43So if you haven't already, grab that beautiful coffee, and let's dive straight out. Now the Hermes AI operating system actually connects your Claude code operating system. And I'll be honest, this is one of the coolest things that I have actually built, and it will blow you away when I show you the detail.

00:59But let me just start by explaining why we're using Hermes or Hermes. We have that's we have some dispute on how we actually pronounce this and Claude Code. And why is this something that everybody who's ever using Hermes actually needs to get their hands on?

01:11So first thing about Hermes is that it has gained 60,000 stars, okay, in the last couple of months. It's way over six figures on GitHub, which is just Internet updates. We like this, basically.

01:21To unfold it, it's yours, and it's got loads of gateways. It is an agent that dreams, that thinks, that self improves. That's why it's gotten so much hype.

01:29But this leads on to the huge problem that we have. And it isn't just Hermes. It isn't just OpenClaw.

01:34Any individual AI system has no handshake. So the two ways that we do work with AI at the moment that are trailblazingly getting better is we sit down and we use codecs or we use Claude code.

01:47That is this guy over here sat at the desk. Right? He's coding.

01:50He's building applications. He's talking to Claude on his computer. He's talking to anti gravity.

01:54Awesome. Then over here on the right hand side, we have our AI assistant, our AI agents. We're talking to on Telegram.

02:00This is Hermes. This is OpenClaw. This is any different AI assistant.

02:05And whether you're using GravityClaw, OpenClaw, ClawClaw, any of these things, what I'm gonna show you applies to everything. Now the issue is just no handshake.

02:13In other words, the stuff we do over here never connects to the stuff that happens over here and vice versa, which means that you never actually have an AI. Any AI system that has a full overview of your entire world. And that means that you lose complete context everywhere you go, and that's just the beginning of the the issues that we have with this.

02:31So you have Claude that knows about your repos, your ND preferences, bug fixes you run at 2AM, and all the software and thoughts and conversations you have with it. Then you got Hermes that knows your brain dumps in Telegram, your conversation. The problem is that you mess you know, there's a Hermes on the walk, you open Claude code, you type it again, it's never seen again, which is what the idea of the Hermes operating system that actually fully connects to the Clawd Code operating system actually solves this problem and unlock some capabilities that you cannot actually use otherwise, which is freaking incredible.

03:02And I'll show you why that is. So you think about this, Hermes hears it, ClaudeCode knows it, and vice versa. So Hermes itself will know everything that you've done on ClaudeCode as a shared memory and vice versa, and it also unlocks some really cool capabilities.

03:16And again, the same thing happens whether it's OpenCLAW or any different variation. We're gonna be using Hermes in this video. Now I'm gonna show you exactly how to set up your Hermes in three steps with this beautiful system and all its new capabilities.

03:28But to do that, we have to understand what this incredible visualization layer is and what that enables us to do that we couldn't really do before.

03:36So this is the Hermes operating system. It actually connects to the ClaudeCode operating system.

03:41I'll put a link down below so you can check this one out. The Claude code operating system shows your AI spend across every single model, DeepSeek, ChatGPT Claude, gives you a breakdown of your usage.

03:51It actually dreams for you overnight. So based on every message you've ever had with Claude, it will give you improvements about, hey, Jack. You're reaching 87% of the work to OPUS 4.7.

04:01Most of it's Hiker territory. So I actually built a system here where Claude will dream for you overnight based on all this information that happened on your computer. And you'll see why this is so important for Hermes in a second.

04:12Because not only does this have every model you're using, all your chat logs from any AI using your computer, it has all that usage data. It has these beautiful independent memory systems.

04:22Okay? They're all connected together, all your connections, and all this data and your usage.

04:26Now what's cool is we can actually leverage that in Hermes. If I go over to Hermes agent on the left hand side here, you can see it shows me all my connections. It tells me the version I'm on.

04:35It tells me what model I'm currently using, memory, my weekly streak, and I have these wonderful chat windows. So for example, here, I can just chat with Hermes if I want to in the browser. But this is where it gets crazy, is you have this thing here called the Pantheon.

04:48Now the Pantheon is essentially these custom AI personas that I actually built with Hermes. It's really freaking cool. And we'll get more into what these do in the video.

04:57But effectively, what they let us do is assign models and roles and personalities. For example, let's say that you want to do some deep research.

05:05Right? And you wanna use a specific model for that. You may wanna use Labyrinth.

05:09Or if you want to do some autopilot and cron stuff, or maybe you wanna philosophize and do some deep reasoning, and you wanna pull up a specific model. Well, we can add these and create these here, give it a name, effectively a description, a system prompt, tell them which model you wanna use, and then give it a a name and and brand, basically.

05:25And I can even add these personas by clicking this button here, selecting who I'd want. So let's say, for example, I wanna go with the Oracle. I can describe the name, the description, assistant prompt, and model.

05:35And, you know, let's say that I wanna do loads and loads and loads of deep research overnight. I may wanna use DeepSeek model or a free model. I can actually specify this, And what I can do is share this then with Hermes agent.

05:46And effectively then, if I say, hey, go and do x or use a philosopher or use Mercury or Labyrinth, it will do that for me using the models, the systems, and the prompts without me touching a single thing. So we build out this pantheon of different characters and and models it can use.

06:01And again, this is a real simple system. I'm gonna show you how we run this all the way through in the video. Now what's really cool here is all the memory from Hermes is reflected here.

06:08This is a an example. One, we're gonna set up completely for fresh together. We have a user profile, what the agent knows about us at its soul.

06:15We can connect Obsidian to Hermes as well, which is fantastic. And then we've got this Claude OS bridge.

06:21And, effectively, what this does is gives all of the information that I showed you earlier in this section here, uh, on your computer from all the models using, all your knowledge systems, all this beautiful stuff, and brings it over to Hermes, which is fantastic. So we can bridge them together by using this real simple install prompt.

06:36And at the bottom, you can see effectively all of your skills that exist within Hermes and what that looks like, and then a quick little hacks at the bottom. So this is the idea of visual intelligence, a operating system that combines your entire world of connected intelligence.

06:52And that sounds super fancy, but what it just basically means is you have everything now in one place. Now you can build this with all the knowledge that sits within your computer and Hermes. I've also put a link to all of this inside the community, the ClawCode operating system, the Hermes agent.

07:05You can literally just go ahead, grab it, and then you're, like, running with it, basically. It even has a really cool setup wizard that runs you through everything, gets your name, gets your profile photo, finds all of the apps that you have. For example, it will find for you all the things that you've got, And then you can go through the whole setup, and it's all kind of, like, synced up.

07:21So it's like plug and play. I actually took me seven hours on the onboarding alone just to get this set up, just to give you an idea of, like, what goes into building this, but you can do all this stuff with everything on your computer. But the bottom line is you want some kind of beautiful visual information dashboard to connect your worlds together.

07:35So now we understand the visual intelligence layer. The next step is to actually install Hermes onto our computer so we can begin connecting everything together.

07:43Now I'm gonna come up here and grab on Hermes. Now the way we get this done, let me just out and exit the demo so I can show you this. Now I did a video explaining that you can essentially connect everything together while we would leverage bits of Hermes, bits of OpenCLORE, and I do believe that is the best way to get a custom AI agent model.

08:00The reality though is that this is the fastest way to go from zero to one, to get a working agent in the system using Homebase, and it's very good. But if you want the best of the best, you're gonna wanna leverage different aspects from different softwares and build it as you go. I'm do something in the community on cost building up those custom beautiful agents, but just so you understand how this fits inside the matrix.

08:19So what we're gonna do is come down and grab this code here. It's also available on the Hermes website, of course. And you can even just say to Claude, hey.

08:26Install this, please. So we're gonna head over to antigravity. And if this sounds like I'm speaking Spanish, check out this master classroom screen.

08:32I run through everything, and I won't feel when you come back. Welcome back. So now we got the terminal.

08:35All we're gonna do is come down here and paste in the code, which is fantastic. Of course, in antigravity or Claude or whatever, terminals at top of the terminal, It's going through and testing everything, which is great. And then literally from this, we can just begin the full setup journey inside this terminal window.

08:49Okay. Beautiful. So once we've entered in the code, we can pick everything we want to.

08:53I've chosen OpenAI Codex. If you don't have that, by the way, it'll just open it up and run through. We're happy with this one.

08:59So we're gonna have to click enter on that, which is fantastic. This is good because it's free with your chat GPT subscription. So that's decent.

09:05The other one that we wanna have in there is Openreach, obviously, but we'll start with Codex for now. Click on enter, which is fantastic. Then we decide gonna use our existing credentials, which is great, or we can just re authenticate.

09:15So for us, our existing credentials are fine. GBT 5.5 is awesome. Our text to speech provider, we're happy using the current one for now.

09:21Terminal background, we're gonna keep it here and run it on our beautiful home computer. We're not gonna run this one on a VPS. This is gonna be fine on our home MacBook, which is running all the time.

09:30Click on current. Max iterations is 60. That's fantastic.

09:34Then we click on all for this one. Compression threshold is fine. 0.8 is good.

09:38It just means the closer it is to 0.95, basically, the less often it compresses. If it compresses too much, it can be a little bit more forgetful.

09:46So we're happy with the slightly higher rating there. And then for session reset mode, we're gonna ahead and keep current settings for that. And then platforms to configure, we're gonna select Telegram and click on enter.

09:55And it's gonna ask for Telegram bot token. Now we use Telegram because it is a gorgeous API. It's so easy to use.

10:00To do this very quickly, just open up Telegram, download the app, get the bot farther. I'm gonna do forward slash. I'm gonna bring him over here so you can have a good beautiful look.

10:07I'm gonna say new bot. Click on this guy over here. Great.

10:11One new bot. We're gonna call it Hermes bot. Great.

10:13Let's choose a username. So go ahead and pick a random name for that bot so I could call it something like this. And you see, guys, you'll get a token.

10:18All you're gonna do is come back over here and just enter in that token. Hit enter. And then it's gonna ask you for something really crucial, which is allowed user IDs.

10:25What this effectively means is that only the IDs that you give it can message your Hermes bot, which is fantastic, which means unless they have your mobile phone, they're signed in, they can't message your Hermes bot. To So get that, come back over to Telegram. And you're gonna search for account called user space info, drop it any message you want to or forward slash start, and it'll give you all the information.

10:44And you're looking for your ID, and all you're gonna do is literally copy that ID, come back over, and throw it into the terminal. Then, of course, you can go ahead and confirm your user ID by pressing y. Once you've done that, it will ask you if you want to launch it as a service.

10:56Just say yes. Then you're gonna come all the way down here, for example, hit on done, and then we are finished. Then guys, you can actually come back to the bot further.

11:03Click on the link to open a new tab, and you should see your Hermes bot here. So I can click on start. So now we can actually begin our conversation.

11:09I'm gonna say, hey there, like so. Drop this guy a message and see if he comes back to us. He's typing.

11:14He's having a think. We can have a little a little play around with what he's gonna say, and this should be Hermes bots. And look at this guy as he's come back.

11:20Hey there. I'm Hermes. I can help you with anything.

11:22You're fully ready to rock and roll. And Hermes is officially downloaded on your computer, which means that you can use this on your phone wherever you want to. So we land into the dashboard, which is awesome.

11:31We can see all of our connections here and our global universal AI intelligence. Obviously, if I click on Hermes, I can see all of Hermes integrations, which are looking very lonely right now to be found. We've got the model.

11:41We've got the active model that we're using right now, which is gbt 5.5, and then we've got what we use this week. Now here we've got a chat window, which is very, very cool.

11:49And effectively, this shows us all the conversations that we have. So this shows that, hey. There's a conversation on Telegram.

11:55You can just talk to it here if you want to. Obviously, it's part of the operating system, or you can talk to directly within basically, you can use it within Telegram if you want to. So you see if I bring Telegram up, for example, and I give that prompt, something like this.

12:05Hey there. I would like you to connect to GitHub. Let me know anything you need from me in order to connect to GitHub just so we can start doing some very cool things.

12:13Paste that one off like so. Send that off. And what you'll see now is this will now appear for us actually within our own our own desktop environment and also of Hermes intelligence system.

12:21And interesting, guys, it's come back and it's on a skill view. Gets it both. It's checked out Hermes agent.

12:26It's checked all the different versions. And because I have this connected at part of the CLI, our command line interface, it's already connected.

12:32So I'm actually gonna update the Hermes bar here to include all of your CLIs, uh, as we speak. So now we've got this. I might say something like, hey, dude.

12:39Create me a brand new repo called Tango Blast, please, and let me know the link to that repo once that is complete. Send that one off just to validate it's got the skills.

12:47Send that one off, and you can see him is working in the background. And, dude, look at this. It's created the repo for us literally in the chat.

12:52I want you to believe, most importantly, we now see GitHub in the terminal, which is in our command center, which is great. Then we'll click this guy right here and open him up, and we can see this is Tango Blast just been created. Fantastic.

13:02So now we're lock down. The next thing that I want us to do now is take a look at building out persona. So this is really important because if you're doing certain tasks, like, we don't need Albert Einstein to wash off to mop our floors.

13:13Right? We want Albert Einstein scribbling, not making much sense on a whiteboard in a very esoteric area practice that only five people can understand. So to do that, we're gonna come down here and install the Pantheon, is very cool, by clicking on this guy right here.

13:24And then it should pop up and we can begin it. So we've got labyrinth. These are a couple ones I thought were really cool just to get started if you just wanna get an idea of what it's like or you do let's say we got the philosophy here, which is fantastic.

13:34We've got a description for wrestling with ambiguous problems, puzzles and threads, questions, premises, etcetera, etcetera. You are the philosopher. Treat every question as a starting point, not an instruction before answering surface and matter question behind the question.

13:45I think that's very, very freaking cool. We can add a persona here if we want to, of course. Say we like the, you know, Orpheus or whatever.

13:51We can add various different things that we like. We're gonna go ahead and cancel that. Now what we wanna do here as well is back up, Hermes.

13:57So to once we built out all the different personas we want to, see maybe one form research, one for tool calling. So anything that's doing deep research, like for your morning briefs, I want you to delegate that to a free model.

14:08Like, absolute like, you don't need your strongest models doing that, which is why, for example, autopilot and cron, I like to use Mercury because it just does a lot of stuff for me in the background. So what I'm gonna do now is I've got this stuff. The first thing that we need to do to activate the Pantheon is take Hermes anywhere.

14:24So look at this. Mirror your Hermes for private GitHub repo, so your config and personas survive a machine swap. Every edit is versioned, and you can roll it back if something starts misbehaving just in case it goes wrong.

14:34Two prompts, paste each one into your Hermes telegram chat or any Hermes session, and she'll walk you through the rest asking for what she needs. Beautiful. So connect the Hermes to GitHub.

14:42Again, just copy paste this one. We can throw it in the chat window above, or we got Telegram here if you want to. Come over here.

14:47You can come down and literally drop that bug by in and let it run wild and do its own thing. Then it'll start to think and effectively just create this GitHub repo for us. And it's really important because one of the things that can happen when you're doing all these different features is it can start to, you know, forget things.

15:00And just like that, it's come back. So absolutely, I went for your confirmation before creating repos and copying files. What I need from you, it's got the Git CLI already.

15:07Git username, it's got that. Repen name is awesome. Backup.

15:10I'm gonna say, sounds great. Go ahead and create all that for me, please, and let me know when that's done. And then I want you to set up a crunch schedule.

15:16Let's say, I don't know, 11PM every day. You're gonna do a full update for me, please. Beautiful.

15:21So now we have this thing. We've created a mirror of this in GitHub, which we can use anytime we want to. The next thing that we need to do then is basically go ahead and recreate everything that we just made in the Pantheon.

15:31So all we're gonna do here is to paste this into Hermes. After we connected the repo, she'll push the latest persona of animals, labyrinth, mercury, philosopher, and anything we've added. And we're basically explaining where this is, what it looks like, and how everything's sort of connected.

15:43So we're gonna copy and paste this like so. We're gonna copy that, come over, drop that one in, and then we're ready to rock and roll. Beautiful.

15:49And now it's come back. It's actually found everything there. Now here's the cool thing, is the way that I've set this up, guys, is that when you add these and if you build yourself, I recommend you do the same way as well, the effect of what this means is when you build these different personas as when you change them in your operating system, it's automatically changed within Hermes because the way this works is these all live in a folder.

16:09So if I add new files to that folder, Hermes knows where to go in the folder. So if I change the job of the philosopher, he does a great job, but he already changed the job. Same with Mercury over here.

16:18I add new ones. That's all gonna be updated in the same location, so it just works fine and perfect for us anyway. And what this saves us doing as well is repeating the same thing a gazillion times.

16:26So I'm gonna say, that all sounds great. Go ahead and do that for me, please. Beautiful.

16:30And just like that, we're all ready to rock and roll. And now when we add new different, basically characters and players to our pantheon and gods to our pantheon, they'll be ready to rock and roll.

16:38And then for example, guys, let's say I want to invoke Labyrinth. I might say something like, hey, that go ahead and use Labyrinth and do some deep research on the best strategies for email marketing, and I would like five of the best headlines for a roofing company, please, when trying to get new clients.

16:54Send that one straight out. And as you can see, guys, as it's now delegated the task, and you can see now are basically Labyrinth, Mercury, and Philosopher are fully synced, which means it's linked and connected now to our new agents. Beautiful.

17:04And we've summoned Labyrinth. Labyrinth has gone ahead and done this. And, of course, if we're using OPUS 4.7, it would have gone ahead and used a different model to do that for us.

17:11And look at this. Catch storm water with a dream demand, offer a low friction next step, segment by homeowner situation, lead with trust proof is given us good details, and it's gone ahead and created that for us, which is fantastic.

17:21And now we've covered exactly how to connect Hermes to anything. That's great. But now we need to take this to a completely new level.

17:27And if your system doesn't have this information, if your Hermes agent doesn't get this, it is not performing to its potential. We're gonna give it an incredible memory system to Obsidian, and then I'm gonna show you how you can supercharge it with an incredible skill that most people haven't even heard of.

17:41Beautiful. So the first thing I wanna do now is connect is agentic memory. I covered this in the full course I did inside my community that goes foundation sets up all the way down to power features, memory systems, this system, and loads of other cool stuff like monetization are pulling down below if you'd love to check that one out.

17:57But the cool thing here that we talk about a lot is this Obsidian system run. Now if you have Obsidian installed, it looks a little bit something like this. You can see I've got my memory core here and it shows me everything I've got, my desktop Obsidian.

18:08Really freaking cool. I just get to visualize all my data, which is fantastic. Now what we can do first of all is come into Hermes.

18:13And you can see these are the memories that it's starting to gain for me. I wanted to install this fresh view on a desktop. You could see user has Hermes Pantheon.

18:20As you can see, this is the agent memory. This is the user profile. And then over here is the soul, which we can build out shortly.

18:26Now what we wanna do is just clarify that you have a vault. So come down here and click, I've I've got a vault. Then what we're do is let you come down here and you're gonna copy this code right here.

18:33And if you don't know the exact destination of your vault, you can just ask ClaudeCode by copying this and throwing it into ClaudeCode. So here, for example, I've got it here.

18:41I'm just gonna gonna paste that in and I'll probably just say something like, hey. I was doing a deep YouTube strategy. Just find me the focal location, please.

18:47Just like that and you can send this one off. And then we've got it here. So we just copy this.

18:50It comes straight back over. Throw in the path like that, is fantastic. And say I've run it now, confirm, and said Obsidian is connected.

18:57Just to double check this, we can literally come back up and drop a message over to Hermes agent if we want to. And as you can see here, for example, I said, hey. I've added it.

19:04I'd like to reference this Obsidian Vault when answering questions, please. Here is the file location saved. This is memory.

19:08I gave it to it And it's saved. I'll use this wiki as your Obsidian vault and reference search it where my questions may benefit from the notes. So now we've connected Obsidian to Hermes.

19:17Awesome. Next thing we wanna go ahead and do then, guys, is shoot over to the home section. And you come to the home page.

19:23You can see now that all the conversations you have with Hermes are now fed into your main operating system. So if you're using a Claude code operating system, you can actually have all these knowledges when it's giving you these really cool recommendations that has that full level context when it's making decisions, which is really cool.

19:38Then if you come back over to Hermes over here and then you scroll down, well, you can actually see the bottom as this really cool thing. Let me go ahead and grab it for you. It says, it's Claude OS bridge.

19:48So this lets Hermes read the dashboard on request and ask what did my dream say or what's inside my Claude operating system. It basically helps to understand because if you when you use Claude code in different models, it actually saves your chat logs.

20:00So what this lets it do is query that and think about that and dream and think about, well, actually, Jack's been talking to Claude Code a lot about these kinds of topics. Maybe I should bring this in or there's a really key piece of context over there that it never mentioned to me. So we effectively connect the intelligence together, which is awesome.

20:17Good old full prompt. Basically, tell the oh, you can kind of explain it, like, what you're trying to do if you want to or copy a prompt to complete your call. It just makes it a lot easier to do.

20:25Then effectively come over to the chat. I'm gonna drop this bad boy in right there, send it off, and let it do its work, or basically explain, hey. I want you go through my entire computer.

20:32I'll find all the stuff that's relevant that could be helpful in doing this, and I want you to set up a cron job that every night you're gonna think proactively about ways in which I can improve based on all these different conversations and things like that. That's effectively the next prompt, which is gonna be like this.

20:46Hey, though. Based on all the information that you have in our conversations and also that you glean from the computer based on these files, I want you to delegate a cheap model to reflect or use GBT 5.5 on ways in which I could potentially improve and be better and come back in the morning with a couple suggestions of that I could potentially action two to three max.

21:06Do that at, let's say, 8AM Dubai time alongside a beautiful morning brief for me. And just like that, it's come back and done it. It's set that cron job for us.

21:14It's gonna use the metal llama model. I don't mind it using the chat g p t model actually because we've got it all in our subscription. Hey, dude.

21:20Just do me a favor. Actually, use my chat g p t 5.5 subscription now, please. I think that should be fine for this particular use case.

21:27Awesome. Send that one off. And just like that, now I'm just every morning, gonna be thinking proactively based on all the conversations across everything, what we want to do.

21:34And this very nicely takes us onto some beautiful use cases. So we know loads of the classic ones like, hey. Why don't you go ahead and set the morning brief?

21:41We get that. But let's talk about something that's gonna grow your business or grow your client's business. Let's say that we're out and about.

21:47We're working with Telegram. Now we have this fantastic connected in Internet ecosystem of things. Why don't we look about getting leads?

21:53And before I double click on what that would look like, I just wanna confirm what we've done here. Right? We've got Hermes, which is great.

21:58We've downloaded that. Second thing we did is we got the Hermes personas. We connected it.

22:03We did loads of beautiful things. And the third thing we did is we built a beautiful skill together. And one other thing that we can do before we actually build the skills out is to come back over to the Hermes agents and effectively can say, hey there.

22:15I would like you to build a soul.md. I'm correct in assuming that your soul dot m d basically is all the key information about me. So I'm gonna brain dump some information, and I want you to retain that in our conversation so you can give me better results.

22:27Is there anything you'd like to know from me? And then literally just brain dump everything to it, and this will then appear in your Sold. M d, and you can track that.

22:35So one of the skills I wanted to take a look at here is gonna be a lead scraper, and we're gonna go ahead and use Apollo. Apollo's pretty much the gold standard when it comes to actually finding people.

22:44So the best ways to grow your business is to find people that can help you. This has got a gargantillion features that we could just go and use.

22:52But what I'm most interested in here is actually the API because sometimes we just have ideas of you know, you'll find it yourself when you're having coffee. Right? I wanna target that person, but maybe it's not convenient to have your desktop out and start building campaigns.

23:05Maybe we just wanna check to Hermes and say, dude, go and find these email addresses, find out who they are, and, like, set create create a campaign for me and get this business. And we can now do that. That's one of the coolest features because the delta between idea and average should be small.

23:18And when you know how to connect things like Apollo, you can connect anything. So to grab this, we're gonna have to grab your API key from Apollo. Great.

23:24So to find this, come down here to admin settings right at the bottom and then click on integrations. When you're on integrations, we should be able to find API key, but you can come down here and type in API key. Apollo API.

23:34Fantastic. Come down and click on connect. That'll open a brand new page for developer.apollo.io.

23:40Fantastic. And on the left hand side, click on API keys. Cool.

23:42So I've laid my old keys so I can show you a brand new fresh one. So click on create new key here. I'm gonna call this one something like Hermes.

23:48I always like to name the thing that it is. I'll say, for use when using Hermes, whatever the thing is, come down, select API keys. We're just gonna set this as a master key.

23:56Come down and click on create API key. Beautiful. That's done.

23:59I'm gonna copy this one. Superb. And then what we can do is let you come back over now and connect that with Hermes.

24:04If I bring up Telegram, for example, I might come down here and say, hey there, dude. I would like to do some automations with Apollo. So I've got the API key.

24:12Let me know everything you need to do to add this as a connector and a skill, please. And so the easiest way that this is just to do it, which is freaking cool, is you can literally do a bash command, which is this space your API key. So let's go ahead and run that one together by opening anti gravity.

24:26I'm gonna come down here and get a new terminal. By clicking on terminal, click on new terminal, and you can literally just ask Hermes for step by step instructions of how to add your API key in. And it's good to add it to environmental variables like this because it will be stored in the chat logs, which will be in GitHub and also at other places.

24:43And obviously, model reads it. You can throw in. Of course, you can.

24:46It's up to you if you want to. But if you just wanna be absolutely eyeing cloud, you can add it in using environmental variables if you like. Beautiful.

24:52So once you've gone through that process of connecting it together, we can also do some really cool bits and pieces. But first of why are we using Apollo? What does that make sense?

25:00How does this all work together? Well, a couple of things that are good to know about this, why it's worth taking a look at.

25:05They have the biggest b to b database, which is cool. They have emails that will get delivered. Mobile numbers are pretty decent.

25:11Intense signals and scoring so we can see whether or they wanna buy. Of course, you can go and scrape things using different services. But if there's a reliability factor, there's a size factor, there is actually hit rates.

25:22What people don't appreciate sometimes is that when you scrape a lot of data, that your hit rate can be quite low. Like, you can only hit sometimes, like, you know, x percent of them rather than as many as you can. So we want good reliable databases.

25:33And if you know you've a great product, it's really cool. We can do sequences and cadences all within the platform, which is really cool. So you don't have to buy a second subscription, which is decent, and it can play nice with your stack, which is good.

25:44So but this is continued refreshed. It's one API, one quota. Compliance is baked into it.

25:50Intent and scoring file happen automatically, and we get the sequence and the dialer in the same screen if you wanna take this to a level on the platform. And then scraping your own obviously has things like GDPR risks, your intent signal.

26:00You may still need to get sequences and that kind of stuff. But let's go ahead and test this together with a good use case because you've seen exactly how to connect to Hermes now.

26:08Let's put this puppy to the test. So let's try this one for example. Hey, the dude.

26:11I've been thinking recently that I'd quite like to target roofers because roofers are high margin, and it's a bit of an unsexy business. So I'd like to see for me a bit of a prospecting brief on roofers using the Apollo skill. And then on top of that, see if you can find for me, I don't know, 20 different roofing companies in Austin, Texas that I might want to reach out for.

26:31Just give me, like, the business name, a little bit of detail about them, and then I can start thinking about whether or not I wanna get contact info and what that looks like. Thank you. I'm just gonna send that off back up in the voice that that could be anything, And then we can come back and get this beautiful this beautiful information to help us grow our business, basically.

26:46And we've got great pain points for roofers here. Past customers are not being reactivated after storms. Leads come inbound, but they don't get followed fast enough.

26:53Really cool. And we've got a strong offer. Freaking awesome.

26:57Then we've got 20 Austin roofing companies down here. We've got Kidroof. We've got Texas Fifth Wall.

27:01All these different guys. I've not included the email address down here, but you can see we've got all these different messages now using the Apollo API. Look at this.

27:08My top average picks, if I were prioritizing first, I'd go with a b c d e f g. And guys, we can connect this now to Gmail to draft emails and send emails to these people, which is amazing. One of the best ways to do this, by the way, which is just so easy, is something called Zapier.

27:21It's like a absolute cheat kit for connecting things. So you're gonna come up to this website. On this MCP, which is great.

27:26Now I would recommend this. Never give OpenClaw or Hermes the ability to send emails only to draft.

27:33I'm still at the point where I would never let them run riot. I'm a couple of my buddies have done that, and they've been a little bit sorry about that. We're just not quite there yet with writing, but drafting is fine, and you can approve it, which I think is awesome.

27:46You So can pick the popular AI agent that you want to. You can come down here, see all of them. It's really freaking cool.

27:50We're gonna come down here to the new MCP server. And what you're gonna come down, Clipman, guys, is Open Claw. Don't worry about the fact that it says Open Claw.

27:55We're just gonna substitute that for something different. And instead, you can copy the prompt here and then come back over here, paste this one in, and then instead of Open Claw, just come down and change this one over here to Hermes, and then come down here and just make sure it's fine and then get ready to rock and roll and just send that one off.

28:10Now when you click on this on the left hand side, you can see we can now tools. So what we're gonna do, add a tool. I'm gonna come down here, find Gmail.

28:15And crucially, I'm not gonna give it the ability to send emails. We wanna follow something called the principle of least access, which is that any tool connection has the fewest possible tools available to actually do the thing.

28:27So find emails is fine. Get attachments is fine. Add labels is also fine.

28:31It's even doing it right now as we speak, which is cool. So we're gonna say, hey. That sounds freaking cool.

28:35We're gonna allow that. That's absolutely fantastic. We're already authorized.

28:38Thank you very much. We appreciate your hard work. Archive emails.

28:41Delete. We don't need it to have delete, but I think create draft and draft reply are fine. Create labels is okay.

28:47Send emails is a no right now. Remove labels is also okay. But, again, we're not going too much of that.

28:52Gonna come down and just click on connect. Now this only has access to the stuff that I'm okay with. I say add seven tools.

28:58The other one that we're gonna wanna add is calendar access as well. So that's cool. Let's just come down and give my calendar access also.

29:03I'm gonna get it built to define events, retrieve events, find, you know, busy periods in the calendar. That's fine. Find calendars not required.

29:11Add attendee to event. That's also okay. And I'm gonna give it the ability to create a calendar event since that's not necessarily too messy, and I'm gonna kind of okay with it doing that.

29:19And also deleting an event is also cool. So basically, to your own preference, basically, build this one out. This one's on my computer at home, so unless someone comes to the store right here and they're going to they're not gonna be able to actually physically get the stuff.

29:31And so just give it the access that you're happy with. And when you're happy, come down and click on connect. And Zapier is pretty much like the Willy Wonka's Emporium for connectors and stuff like this.

29:38It's very cool. I'm gonna come down and select my account, which is cool, which is this one, then add my eight tools. And when that's done, I click off.

29:44And then basically, you'll see if I come down here onto connect, I know you'll see an apps that is all physically connected together, which is wonderful. And then just like this, guys, it's now confirming that everything's connected, and that's as easy as was. I can say something like, hey there, dude.

29:55Could you tell me, for example, the title of the next calendar invite that I have, the next appointment on my calendar? Now Hermes has got this.

30:02We'll just see what else to say. This is the easiest way to connect to Google. Like, it it that I'm aware of.

30:06It is just super duper easy. It's fantastic. And look at this.

30:09We're even seeing a preflight compression. So these are tuck ins route. This is a threshold.

30:13Now it's gonna compress a little bit, and it gives us a wonderful answer. And whilst we wait for Hermes to come back with the answer for that, just wanna touch on some pricing stuff on the Apollo side if you do wanna go out that one. It is a paid subscription.

30:22There are free alternatives out there, like, can use Apify. This one is paid for a reason because of the quality, obviously. So all the way down from $0 up to a 119 if organization.

30:33Obviously, you can pick the one that you like. You can use this API from free, but they basically, there'll be certain things you can't get like emails. So just be aware of that when you're actually going ahead and using it.

30:42Alright, guys. And just like that, it tells me my next calendar invite is way in. It does not know about the beautiful nandos I had tonight.

30:49But now we've connected all this stuff together, it just brings on to a very interesting question. And that's how do we exactly solve the second part of this puzzle? We get the fact that we can use Hermes on our phone, which is fantastic, but it's only half of the knowledge system.

31:02And so the next thing we're gonna do is learn how to build this agentic system for Claude, which we're gonna do in this video right here.

The Hook

The bait, then the rug-pull.

Jack Roberts opens with a promise that sounds inflated until he shows the marketing slide: two AIs, zero handshake. Claude Code knows your repos and 2AM bug fixes. Hermes knows your Telegram brain-dumps. Neither has ever met the other — and every context switch forces you to re-explain your entire world from scratch.

Frameworks