The argument in one line.
Most Hermes Agent users drive a Lamborghini at 10mph because they never unlock its memory systems, background tasks, cron scheduling, or goal architecture.
Read if. Skip if.
READ IF YOU ARE…
- Hermes Agent users who set it up but are only using it as a chatbot and haven't touched memory or scheduling
- Builders who want a compressed walkthrough of every major Hermes capability without watching 100 hours of content
- AI power users comparing personal assistant agents and wanting a concrete demo of what Hermes actually does
- Founders who want to run background AI tasks and cron jobs through a personal agent layer
SKIP IF…
- Anyone not using or not planning to use Hermes Agent specifically — lessons won't transfer cleanly to other tools
- Viewers who need beginner AI orientation — this assumes you already know what an AI agent is
The full version, fast.
Hermes Agent has a memory system, background tasks, cron scheduling, model switching, and a goal hierarchy — but most users treat it like a basic chatbot and miss all of it. This 23-minute condensed guide walks through each major capability with live demos: a memory.md file that retains full conversation history and person cards with tone preferences, a soul.md that locks persistent personality and instructions, background tasks that run autonomously without the app open, cron jobs for scheduled recurring actions, and a goal/super-goal system for long-horizon objectives. The framing throughout is that the tool is already powerful — the gap is in how deliberately you configure and direct it, not in what it can do.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:00
01 · Hook and Lambo frame
Opens on custom landing page. Sets shame hook: downloading Hermes and using it like a chatbot equals driving a Lambo at 10mph.
01:00 – 02:03
02 · Memory System
Explains memory.md, peer cards, fuzzy index, 1hr prompt cache. Shows the It never forgets memory engine UI live.
02:03 – 02:44
03 · soul.md
Describes soul.md as the context manual for your life. Fill it out so Hermes knows who you are, how you think, and how you want it to behave.
02:44 – 03:25
04 · Obsidian integration
One-click folder path paste connects Hermes to local Obsidian vault. Live demo: asks Hermes to surface a YouTube strategy insight from stored notes.
03:25 – 04:23
05 · Chat your meetings
Connects Granola via MCP. Asks last meeting title and key point live. Works with Granola, Fireflies, or Fathom.
04:23 – 05:16
06 · Email integration
Zapier MCP for Gmail. Fine-grained permission selection: can draft, archive, label, cannot send. Shows permission UI live.
05:16 – 06:40
07 · Background tasks
/background slash command stacks parallel prompts that never collide. Two simultaneous research tasks running independently.
06:40 – 07:55
08 · Cron job and dreaming
Sets up an 8AM morning brief with a dreaming pre-task at 6AM. Hermes reviews all context before the brief fires to surface 3 proactive daily recommendations.
07:55 – 08:54
09 · Bridge to Claude Code
Claude Code OS bridge gives Hermes full bidirectional awareness of everything happening in Claude Code. Full context enables better advice.
08:54 – 12:17
10 · Goals vs Super Goals
/goal persists for 20 messages as a north star. Super goal adds human-AI handshake so tasks requiring the user are explicitly flagged. Live demo: Launch Hermes Course to 500 signups broken into 6 chunks.
12:17 – 14:00
11 · Switch models mid-chat
/model command. Grok 4 OAuth for real-time X search. OpenRouter for every model. Recommendation: Claude Opus 4.7 for chat quality, Sonnet for most tasks.
14:00 – 15:06
12 · The Pantheon system
Custom agent library assigning specific models to specific missions. Reasoning tasks get powerful models; cron tasks get fast cheap ones. Tag the best model for the best job.
15:06 – 17:16
13 · Anti-Gravity CLI
Gemini CLI successor. agy command, OAuth to Google. Best for multimodal: video analysis, images, long documents. Use Gemini for multimodal, ChatGPT for code review, Opus for design.
17:16 – 18:45
14 · Mission Control OS
Custom-built Claude Code OS dashboard: usage stats, model connections, long-term goal tracking (chief Wiggum), onboarding flow that identifies all apps on your machine.
18:45 – 19:55
15 · GitHub backup
Daily snapshot of Hermes state to a private GitHub repo. Portable across machines. Every skill, soul.md, and memory file backed up automatically.
19:55 – 20:54
16 · 10 key commands
/goal, /handoff, /background, /cron, /steer (direct without interrupting), /resume, /kanban, /curator, /stop, /model.
20:54 – 22:21
17 · VPS vs local deployment
Recommends local (Mac or laptop, 24/7). VPS adds security surface. Docker as isolation option if you want file access control. Local is simpler for Obsidian-connected setups.
22:21 – 23:03
18 · Firecrawl for search
Firecrawl converts HTML to structured data, cuts web scraping costs 80%. Recommended for any agentic internet search in Hermes.
23:03 – 23:20
19 · Teaser CTA
Ends with a next-video tease: the most powerful, underrated connection in Hermes Agent. Classic series hook with no resolution in this video.
Lines worth screenshotting.
- Most people who install Hermes Agent use it like a chatbot — they bought the Lambo and are driving it at 10 miles per hour.
- A soul.md file is the context manual for your life — without it, the agent has no way to understand who you are, how you think, or how you want it to behave.
- Connecting your local Obsidian vault to Hermes lets you query years of your own notes in natural language, instantly.
- Linking an AI meeting notetaker to Hermes means you can recall what was said in any past meeting by simply asking.
- The Hermes memory system uses peer cards — one per person — covering tone preferences and conversation history so it never loses context on who you deal with regularly.
- Background tasks allow Hermes to work on a multi-step problem while you do something else entirely, then surface the results when complete.
- Cron scheduling inside Hermes means recurring workflows — morning briefs, weekly reviews, daily reports — run automatically without any manual trigger.
- Model switching lets you route cheap tasks to faster, lower-cost models and reserve expensive compute for the work that actually needs it.
- The super-goal system keeps the agent oriented toward a long-horizon objective across many separate sessions, not just the current conversation.
- A one-hour prompt cache inside Hermes keeps token consumption low by retaining recent conversation context without re-sending it every message.
Steal the shame hook. Steal the structure.
Creator playbook
The Lambo at 10mph is a reusable template: any tool your audience under-uses has a version of this hook waiting.
- Open with a metaphor that communicates unused potential without shaming the viewer directly. The Lambo does the work before you say a word.
- Structure a 20-tip video as a timestamped chapter guide. Every chapter is a standalone clip and a standalone search result.
- The dreaming sequence framing is novel and quotable. AI runs a pre-task before the main task. Steal it for any automation content.
- End every multi-part tutorial on an unresolved tease. Name the next thing without showing it. Forces the click.
- Screen-share with face-cam PIP replaces B-roll. If your content is software, the interface is the set.
- Build a custom-branded UI like the Hermes OS dashboard. It signals product depth and becomes free thumbnail content.
Terms worth knowing.
- Hermes Agent
- An AI personal assistant application built for power users, distinguished by a persistent memory system, soul.md context file, and integrations with external tools like Obsidian, meeting notetakers, and email — designed to function as a long-term intelligent companion rather than a stateless chatbot.
- memory.md
- A plain markdown file maintained by Hermes Agent that stores everything discussed across sessions, acting as a persistent external memory so the AI can recall past conversations, events, and facts without re-stating them.
- soul.md
- A context document the user writes once and loads into their AI agent, covering who they are, where they live, their preferences, and how they want the agent to behave — the equivalent of a permanent system prompt personalized to a specific person's life.
- Peer cards
- Per-person memory entries inside Hermes Agent that store tone preferences and key facts about specific people the user interacts with, so the agent can tailor communication and recall context about each contact.
- Fuzzy index
- A memory retrieval mechanism in Hermes Agent that finds relevant stored information even when the user's query doesn't exactly match the stored wording — enabling approximate, human-like recall rather than exact-match lookup.
- 1-hour prompt cache
- A short-term memory layer in Hermes Agent that holds recent conversation context to reduce token usage and avoid re-sending the same background information on every message within an active session.
- Obsidian (note system)
- A local-first markdown note-taking app that stores files on the user's own computer. When connected to an AI agent, it allows the agent to search and retrieve the user's personal knowledge base dynamically during conversation.
- Granola (meeting notetaker)
- An AI meeting transcription and note-taking tool that records calls and makes the content queryable. When integrated with an AI agent, users can ask questions about past meetings conversationally without re-reading transcripts.
- Cron scheduling (AI agent)
- A feature that lets an AI agent run automated tasks on a fixed time schedule — such as daily briefings, weekly summaries, or recurring reminders — without the user manually triggering each run.
- Goal / super-goal system
- A goal-tracking feature in Hermes Agent where users define short-term goals and overarching life goals that the agent references during conversations, keeping advice and suggestions aligned with what the user is actively working toward.
- Model switching
- The ability to change which underlying AI model (e.g., GPT-4, Claude, Gemini) powers an agent's responses on demand, allowing users to match model capabilities or cost to the task at hand without leaving the same interface.
- Background tasks (AI agent)
- Long-running or asynchronous jobs the agent executes without requiring the user to stay in the chat — such as research, drafting, or data processing — that complete and return results when ready.
Things they pointed at.
Lines you could clip.
00:45
“Downloading Hermes and not using these great insights is essentially buying a Lambo and only keeping it in 10 miles an hour.”
07:30
“The Hermes agent is the agent that grows with you. The more it knows about you, the better it gets.”
09:08
“The best personal assistant is the one that you are going to use.”
16:44
“Use Gemini for multimodality, use ChatGPT to review code, and for design, Opus 4.7 is still the king.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00Hermes Agent is the best AI personal assistant on the planet, but most people use it just like a chatbot, meaning they don't unlock its full capabilities. And in this video, I'm gonna show you everything that I've learned from spending hundreds of hours in Hermes. So you don't waste weeks of time, build things superior to your competitors, and do so way more profitably.
00:22And if you're new, I'm Jack. I built in Summerlite Tech startup with the gazillion customers. Now I'm building my own AI startups and I just shared the stuff that actually works.
00:30So if you haven't already, grab that coffee and let's dive straight in. So this is gonna give you everything you need to know about Hermes and at the end of this video, you're gonna have a Hermes agent so powerful, you're not gonna actually believe it. Now first thing to understand is that downloading Hermes and not using these great insights is essentially , which is not unlocking the capability.
00:50Now, the first thing we talk about is Hermes memory. It is what sets it apart from all of the other AI assistants, and it's so unbelievably easy to use.
01:00I'll show you how it works, and little hacks that I'm using to make its memory like some kinda . So you've got this memory engine that explains what's going on behind the hood. But there was an update recently inside Hermes that makes this memory super powerful.
01:15So the first thing you wanna make sure you're doing is make sure Hermes updates. So just make sure you're dealing with the very latest version of Hermes so you have access to all of the latest features.
01:24And in this video, by the way guys, I'm gonna cover some of the biggest questions that I've seen on Hermes. I haven't seen anybody answer. It is gonna help you get way further ahead.
01:31So one of the things that we can do is actually ask Hermes what it was doing on specific days. Hey there, man. Tell me, what was one of the things that we were talking about on Sunday?
01:38And as you can see, Hermes can now come back and specifically mention what you were talking about on that exact day.
01:45Now the actual memory system itself has a memory dot m d, which is just a plain markdown file that has everything you're talking about. You have these peer cards, which is one card per person. It covers tone preferences.
01:55We have this fuzzy index and also a one hour prompt cache, which keeps your token limit lower and will just remember some of the conversations that you've had. One thing that you need to add to Hime's agent is a soul. Md.
02:07I can show you what that looks like on this beautiful system right here. So for example, on Hermes, you'll see you can effectively have a chat with it and it will record everything that it needs to know about your soul. Md.
02:18For example, I'll put this link down below for you so you can grab it. You need to, if you haven't already, give it a soul.md. Explain who you are, where you live.
02:26It's basically the context manual for your life so it understands everything that you're talking about and also how you want it to behave and chat to. Do you want it to look around corners? Do you want it to challenge your thinking?
02:38You need to make sure that you actually fill out the information on your soul. The other thing that you're gonna wanna do is connect your Hermes agent to your Obsidian memory system that lives on your computer.
02:49If you don't have Obsidian, you can also connect it to Pinecone if you want to. This is an example. This is just all the different files I have in my Obsidian.
02:55You can do this pretty easily by just like linking it up or sharing that pathway. This is like a a one click paste and go, but that will really elevate the level of your memory system inside Hermes, meaning you can ask it questions.
03:06For example, if I bring up a Hermes agent right now, I can say something like, hey, dude, in my Obsidian memory system, I talk a lot about YouTube strategy. Go ahead and pull me one great insight for intros that will be valuable for me. And this can actually I go ahead and dynamically search my Obsidian memory system on my own computer.
03:23Then it's also really handy to give it access to AI meeting notetaker. Again, that could be granola, that could be fireflies, it could be fathom. But being able to ask questions about your meetings dynamically is unbelievably helpful.
03:35For example, if I get a fresh chat, can say something like, hey there, just tell me the title of the last meeting I was in and just one point that I made in it that would be interesting or valuable. And then we can send that one off, and then just chat to our meetings on the go, which is unbelievably helpful when you connect it to one of these services.
03:53And as you can see, it's connected to my granola. It knows that I'm using granola. I find granola best because I can just chat to it.
03:59I did move from one the other services, but it's not sponsored or anything, but I just find that one really helpful. And it'll just give me that information whenever I physically need it.
04:07And you'd be surprised how often you wanna recall on your own emails. It is crazy. And they go, for example, from the coffee call that I do in my community every week, where I answer people's questions.
04:16This was from a YouTuber who had a question and we helped them out with that. Very cool. Then once we've crushed meetings, the next one you wanna do is connect to emails.
04:23You can connect to Google directly, but the easiest way to do it I found, I know that I personally use is Zapier. You can click a new MCP down here by clicking on other. And the reason I like this is because it's very easy to actually select all the bits and pieces that you want to.
04:37It shows you all the things you need to do to connect, And you can effectively just pick all the tools that you wanna go ahead and use. So for example, all you do down here, if I click on Gmail for instance, I can select what am I happy with. I'm happy with adding labels, archiving, deleting, creating drafts, but I I may not want it to be able to send.
04:52So you can dynamically configure all the stuff that you want to come down, connect your tools, connect your world, and then connect that to basically Hermes agent so you can do anything you want to and actually ask questions about your emails. Now by using these systems, you will 10x your memory beyond what most people are even doing.
05:08And I'll leave this on screen for a second so you can learn about how the Hermes inbuilt memory actually works. Well, leads us on to one of the coolest features inside Hermes Agent, which is the ability to run things in the background. You'll notice when I say questions like hey, and then you send another one like why, or you ask it any questions, it basically interrupts it.
05:24So we can get around this by using a new slash man called background. So I can say, hey, there dude, do me a favor. Go do some research on the most untapped industry for AI automations.
05:34Okay? Beautiful. That's the thing that I wanted to go ahead and do.
05:37Now, this is gonna basically if I wanna ask another question, I can do forward slash. I can do background. Okay?
05:42And what I can do with the background prompt is I can be like, hey, actually, dude, tell me also what is gonna be the best place to live on the planet? What has the best work life balance? Whatever the thing is.
05:52Which is basically means you can have many different thoughts running at the same time and they will never collide into one another. And this very nicely takes on to another powerful feature. And by the way, if you wanted to learn about Claude Code and the Homey's agent system in detail, I've got the world's most comprehensive course that I've ever pulled together on Claude Code taking you from zero to pro.
06:09We cover memory systems, the Hermes agent stuff that I've never covered on YouTube, apps, build anything. I get so many messages on this every day. I wanted to make sure that you're aware of it because it is crazy.
06:19I'll put a link down below if you wanna go and grab that and level up. It is pretty nuts. Now, one of the things I really wanna double click on here with the Hermes agent is the fact that we can actually schedule tasks, and this is something that I recommend you do by way of what we call a cron job.
06:33For example, I can say, hey there dude, every morning, I would like to set up a morning brief to fire off at 8AM. And on that brief, I would like the following. What is the what is it for today?
06:43What is a motivational quote from one of the top three entrepreneurs on the planet? What meetings do I have coming up that day? And crucially, one key thing, I would like you to begin a dreaming sequence.
06:54I would like you at 6AM every morning before you send this morning brief to go ahead and actually consider based on all of our conversation history, everything that you know that I'm doing, all my meetings, everything we've spoken about, what would be three core recommendations that you would have for that specific day? And one non negotiable thing that I need to be doing that day based on all the data you've got both long term, short term, and access to all the files and systems that you have for me.
07:22Now this is a really important prompt. And the cool thing about this prompt is I'm now getting Hermes to dream. Now Hermes is the agent that grows with you.
07:32The more it knows about you, the better it gets. That's one of the most fantastical things about this system.
07:38So I'm gonna wake up now with a morning brief. It also, by the way, has access to my entire ClaudeCode system.
07:44This here is the Claude Code operating system. It's got the Hermes Agentic operating system. I'll explain about this in a second.
07:50But the cool thing that we can actually do here is connect Hermes to Claude Code. So effectively anything that I'm doing I see even Claude Code's got this dreaming feature here.
07:59But anything that I'm doing inside Hermes or CallCode can be connected. So my CallCode knows exactly what's happening in with my Hermes, and my Hermes agent knows everything that I'm doing on my computer.
08:12Right? It's fully fully connected to it. You see this CallCode OS bridge right here?
08:15What this basically means, any messages I'm sending to Claude code, anti gravity, codecs, you name it, I now have access to in Hermes agent. Why is that important?
08:25Because we have this split world. Sometimes I'm coding, sometimes I'm just talking to Hemi's agent. This bridges the gap, meaning I can actually have full context.
08:33And how can you possibly give the best advice if you don't have the full context of the thing you're talking about? Now as you can see, it's come down and it's basically explained that entire brief for me.
08:42So now I'm gonna have that, what we call a cron job happening whenever. You can also say, hey, Hermes, remind me in thirty minutes that I need to go take my dog out. And it will message you.
08:50So it is literally treat it like your personal assistant. We like it. We're for a birthday parties.
08:55We're big fan of Hermes, but treat it like your personal assistant. And then this brings us nicely onto OpenClaw versus Hermes versus GravityClaw versus your own personal assistant. The thing is this, the best personal assistant is the one that you're going to use.
09:09Of all the ones I've tested, Hermes agent is the best at getting you from zero to one. I've listed some of the issues that I found with OpenClear. I found that it folds itself into a pretzel when I use it, and I try to push too much.
09:22But you can see some of the key differences in terms of its memory cross sessions, background tasks, mission control, loads of interesting things that I think you'll find valuable. But the cool thing is I wouldn't run six or seven different assistants. I find it best to have one and double click into that and give it all the strength and connections that it Now this takes us on to another really cool feature in Hermes, which is the goal feature.
09:42Idea being that you basically come down, you can type in forward slash goal, and you can give it a task. So for example, I can say, great. What do wanna talk about?
09:50And essentially, the idea here is it lasts it persists for 20 messages. And Hermes will basically go, we'll just keep on working until it solves the problem that you've got it. So goals are a great one.
10:01It's like a north star. Like Hermes, for this session, for our conversation, I want you to focus on this thing.
10:07And you can actually 10 x the forward slash goals by using something called a super goal. Now a super goal is something that I built out because I basically spent days looking at this and I realized that goals are great, but they have one limitation. Mainly that there's no human AI handshake, which means that if you say to Hermes, hey, I wanna go build a YouTube thumbnail for a video that I'm doing, it will keep going and it will give you an output.
10:29Or like, if you you may give it a goal that requires you. For example, in the super goal system and by the way, I'll put the full resources for you down below so you can literally grab the, basically, short goal skill, which helps you write good short goal prompts. And I'm also gonna put down below for you the long goal skill.
10:44So if you haven't got the dashboard, basically what you can do is just use this system. It's basically a mega prompt that you put into Hermes, and it'll it'll basically break down any goal you've got into multiple bite sized chunks. And effectively what it does is assigns actions to either you or Hermes.
11:00For example, shape the course promise. Look, I can click into this. And this has got one big task.
11:05And the goal here is launch Hermes course to 500 sign ups. Cool. So you wanna launch course on Hermes, you love it, you wanna show it to the people.
11:11Great. Well, to do that, there's actually six things that we need to do. And Hermes agent will break those down.
11:16If you use that long prompt, if you're the dashboard, you'll see you'll see something that looks like this. And effectively, it does is simple. Look.
11:21I need to share the course promise. I need to build the opt in engine. I need to write the course sequence.
11:27Once that's happened, Jack, you actually need to go ahead and record the promo videos for it, bro. Interesting. So you see how if we just use the standard goal feature, this never would have happened.
11:36So by having this human AI handshake, we can connect together. This is really good for slightly longer term goals. As you push final sign ups, 500 drops is verified.
11:45And when it's all done, you know, it's all basically complete and you can check everything and it makes perfect sense. And remember, the goal feature is like a Ralph loop. It basically will just keep working and working and working until you define the outputs.
11:57You need to make sure that it's specific. As for example, don't say I wanna grow my business. I want to get a thumbnail that looks like this or I want a document that hits these criteria.
12:06Needs to be specific. And I'll put a link on screen for my full video breakdown if you wanna check that out a little bit more about it. And then this brings us really importantly onto the idea of models.
12:13So one of the cool things we can do in the Hermes is use our ChatGPT subscription. There's a brand new update, which is very cool, which is basically the ability to use our Grok model, again, completely on the subscription.
12:25For example, if I come down here, I ever do forward slash model, I can actually see which model I'm talking to. And as you can see, I'm now talking to x AI Oauth.
12:33So I can say something like, hey there, dude. Do me a favor. Just check X and tell me what are the most viral tweets going live for Hermes in the last forty eight hours.
12:41This is by far one of the coolest, biggest things you can do with X, AI specifically in Grok, because effectively it has access to the entire x database that would be which we'd need to basically grab from Apify or some of the means if we didn't have that. Now, how you actually physically do that, which I'll show you what's working in the background.
12:59If you open up the terminal, which you can do by command space bar and then type in terminal and it will appear like magic. Is what we can do is you can do Hermes space model like so.
13:10And this will show you all of the different models. Now the ones that you're gonna want to have, x o y Grokwoth important, literally you just hit enter and it will go through the OAuth experience, opens a browser, you sign in, then you're fully connected. It's awesome.
13:22The one that you're gonna use most of the time though is going to be you need OpenRooter. This will give you access to every model. Very freaking cool.
13:29We like that a lot. But of course, for your day to day stuff, I recommend OpenAI Codecs.
13:33I like to chat basically with called Opus 4.7. It's expensive and you do have to pay on the API credits. But it just in terms of how it presents in the chat, there's nothing really better.
13:44You can just use OpenAI Codecs, but I don't ever use Opus 4.7. You can use the Sonnet, by way. Sonnet is just as, like, good in many respects to be fair.
13:52But basically, you never wanna be doing any deep research with those powerful models. Because one of the cool things that you can do with Hermes is delegate. Especially if you have something like I built called the Pantheon.
14:03Pantheon basically being the ability to create different tools for different missions. You can do this just by it's basically creating skills. I personally really like the ability to look at things visually.
14:13For example, if I've got Orpheus here who deeply reasons, I might use a more powerful model. But if I'm in like autopilot stuff and cron stuff, I can specify here basically what is the description of the job, what is the system prompt, I can go ahead and do this, and you can leverage it.
14:26But even if you're just talking to Hermes, you can be very specific about what models you want to do what, which is massively helpful. And I can come down here and look, it's coming down.
14:35It's even pushing about the integration. And with Hermes, you can say, hey, dude. I want to use OpenRooter and delegate to DeepSeek v four.
14:42Okay? And just get it to go and do some deep research on what is the best place in the world to live. Cool.
14:48So this is an a silly example. Basically, you can delegate these research tasks to any model you want to.
14:54And that's how you really achieve this, like, insane level of performance without necessarily burning through your tokens. We specifically tag in the best model for the best job, which leads on to another great update. And this idea that we no longer have the Gemini CLI.
15:09Oh my gosh, Jack. What's going on with it? Well, they're deprecating it.
15:12Now, it exists as anti gravity. So what does that mean? Well, I'll you what it means.
15:17Okay. If I come down and let's create a brand new terminal and shut this one down. Effectively, anti gravity is the new Gemini terminal.
15:25So if I open up terminal, what this basically means is that we can access all of the Gemini models or basically the best Gemini model through this. If you come down and do AGY, once you've installed it, you'll see it appears. If you haven't installed this and so you can download the anti grab TLI.
15:39I did a full video on this like literally a day ago. But if you come down and grab this link, I'll put it down below for you as well. Effectively, what you're gonna do is come down and grab the anti gravity CLI.
15:49Just copy that. Literally come over to terminal, enter it in, and able to use it. And effectively, you do Google or OAuth, which is awesome.
15:55Opens up a new browser. You just go ahead and sign in. And then basically, you're fully connected to the anti gravity CLI, and you can now use that within Hermes.
16:02To say, for example, this is doing some work in the background and it's going on too long, what I can actually do is forward slash stop and it kills everything. Cool.
16:09Then I can say, hey there. I want to use the anti gravity CLI to go ahead and quote for me a one page website on selling websites. It looks gorgeous and beautiful.
16:20Something crazy like that. And I might just say use AGY if it isn't aware. Now this can use that CLI.
16:26Now the best use for the anti gravity CLI remember, Gemini multimodality.
16:33Reading looking at videos. It can analyze videos, images, long stuff like that. Very, cool.
16:39But again, like use Gemini for multimodality, use chat GBT to review code, and for design, Opus 4.7 is still the king. But we can tag in different models whenever we need to.
16:49And so now we have this approach, tagging in the best models. One quick thing to call out on X is a million context window. We like that a lot.
16:56It's very, very cool. But I think the coolest thing, honestly, is probably the fact that it can just search X. And if making any kind of content, your ability to have your finger on the pulse of what's working, what isn't is really really valuable.
17:07So that's kind of the shining grace of Grok. Now, this brings on to another really key thing that I found invaluable using Hermes and even called Code.
17:16And that is the idea of having a mission control and operating system. I built this one. I actually spent over seven hours specifically just on this onboarding.
17:26It takes you through your operating system. It basically gets your photo and it actually identifies all the apps that sits on your computer.
17:33Now, you can build this one yourself too. Obviously, if you want, you can grab this one from the link below in the community. I spent a lot of time building it because I knew how valuable this thing actually is.
17:42But one of the cool things that I think is really important is tracking in usage on things, even if it didn't go off, and understanding the value of that. Now one of the things that you want in any operating system is just to understand what your usage is. You want it to go ahead and dream and find things for you proactively.
17:55On the Hermes side specifically, one of the key things is this. Obviously, chatting is great.
17:59I wanna know my memory. I wanna know my average usage. I want to know what models.
18:04And look at this. It's now telling me I'm using Grok 4.3 and I can see which models I'm connected to. That's how cool this thing is and it tells me all my open connections here at the top.
18:13And literally, when you connect here, guys, it will appear for you automatically. I can see my global connections here. That's how freaking cool this thing is.
18:20You can come down. This is where I talk to you about your goals, your long term goals. I call this the chief Wiggum.
18:26Ralph Wiggum was the kind of guy eating crayons in the corner. Mission control here, this idea of long term goal planning is I think fantastic to have a visual interface for that looks like. Obviously, building out your own systems, your own characters is really cool.
18:38And then the ability, which is really important, which I don't see enough people mentioning, is this idea of actually connecting your Hermes agent to get up. The reason we do this, and this is a task I have running every single day in Hermes, is it takes a complete snapshot every single day.
18:51People may not even know what this is anymore. Right? Maybe they're like 16 or something, but taking a snapshot every single day of your Hermes and backing it up in a private GitHub.
19:00We do this because if we change our computers or something like that, we can always take our Hermes agent with us because it's stored privately in GitHub. GitHub's just a fancy word for place we store things. You know, this is an example of a power design skill that I made for Beautiful Designs.
19:16You know, this is an example of the GitHub repo. It's basically private and it stores everything there. And you can just give a prompt to Hermes and it will store it and back it up so you've got it whenever you need it.
19:25And then with Hermes agent as well, there is also the ability to control your computer. In honesty, when I've played around with this, I've not really found myself using it too much. There aren't that many use cases, but it is really cool that you can ask it to do things by just saying, hey.
19:40Use my computer and do various different things on your actual laptop. But I've not found myself going and using that a lot, so I don't think it's, like, gonna change your world. But it's cool, but it's there and it's able to physically go ahead and do that.
19:51Now the other thing that I just wanted to double click on here are probably 10 commands that you can use most frequently within Hermes agent. Obviously, goal is keep forward slash hand off to switch the model persona platform. Cron is if you wanna do that.
20:04But obviously, with a lot of these commands, if you're just using natural language, if Hermes is doing its job properly properly oh my gosh. How many copies have I had today, guys?
20:13If it's doing its job properly, it should know it's doing automatically, but you can spin it up. Clear you know, steer is a really good one. So for example, you can steer it in a direction without actually interrupting its task.
20:23That's really cool. Resume to pick up a different session background. So again, you could actually have any various different things and stack them up without interrupting it.
20:31Kanban for different multi agent swarm boards. Curator, stop is really key, and model just to switch those models manually if you want to.
20:38Then I bring this down to where should Hermes agent physically live. I don't have a VPS to show you on this. It doesn't make a big difference to me.
20:48I honestly think running it on your computer is the best thing to do because I just found that it's easier. Honestly, in what respects, I think it's also safer because if you're in the if you're in a VPS, unless you tunnel it properly, you do all the correct security provisions, you know, there are people that can run, like, attacks on it.
21:06So you can do it and be 100% safe. Of course, you can. I've shown you videos on exactly how to do that.
21:12But it's on your own computer. It's on your MacBook running over there on a laptop. If I was advising my brother or a paid consulting call, which I don't really do many of right now, but if someone asked me what would I do, my honest answer would be, I think it's best to have your own Mac system or your own laptop and run it and have it running twenty four seven.
21:30It's in secured on your computer in your room. And if you're thinking, that's cool, Jack, but I just don't wanna expose it to loads of different files.
21:38Cool. You can use something called Docker. I covered it in my ClawCode Masterclass.
21:42And effectively, it creates and this is the app that you want, is Docker. Effectively, what it lets you do is create self contained containers.
21:50As you might have guessed it, like containers on your computer that run on your laptop, but are completely isolated. It's like having your house as your laptop or your MacBook.
21:57And effectively, what what Docker is, it has a sealed off room that it lives in, this kind of box, and it cannot access anything else on your computer apart from what you have. So you can use something like Docker and just run Hermes in Docker if you want to. But honestly, I remind on my desktop personally, you can run it in VPS if that's what you want to.
22:14But just be aware of your options. You don't have to run this in the cloud. And in fact, there's some things that just adds more complications, like if all of your memory systems and Obsidian are on your computer, you can do that.
22:25If not, you have to connect it in different ways. Again, completely possible. I just think it's more straightforward to run it on your computer.
22:31And then with Hermes, just bear in mind, you've also got access to different messaging platforms, mobile providers, web search, loads of internal back ends, so many different tools. And the other one I would just call it actually is when it's accessing the internet, the tool that I always use to integrate with this is something called Firecrawl, where pages are basically HTML.
22:46Firecrawl can cut our costs down by 80 percent, and it's basically like agentic web search. It's super powerful.
22:52It just always seems to get me the best information I need. So this is what I tend to use for Internet search when I'm going ahead and do any kind of scraping online. But then this takes us on to probably the most powerful, underrated connection that you could possibly make within Hermes Agent.
23:08And that's connecting it to the world's most powerful intelligence and research platform that unlocks new capabilities that most people don't even know exist, and we're gonna learn that and do that by watching this video right here.
The Hook
The bait, then the rug-pull.
Jack Roberts opens with a Lambo. Not literally, on screen, in metaphor. Most people download Hermes Agent and use it like a chatbot. That is driving a Lambo at 10mph. This video is the manual nobody reads.
Named ideas worth stealing.
00:45concept
The Lambo at 10mph
Framing device: buying a Lambo (Hermes) but only driving it at 10mph (using it like a chatbot). Communicates unused potential without shaming the viewer directly.
Steal forany you are leaving money on the table hook for AI tools, software, or systems content
02:03concept
soul.md
A plain markdown file giving Hermes your life context: who you are, how you work, how you want it to behave. The context manual for your life.
Steal forany content about personalizing AI systems or building an AI that actually knows you
06:57concept
Dreaming sequence
A pre-task cron that fires before your morning brief. Hermes reviews all long-term and short-term context to generate proactive recommendations before you wake up.
Steal forautomation content, morning routine content, AI as a chief of staff framing
10:22model
Super goal and human-AI handshake
- Define the goal
- Break into chunks
- Assign each chunk to human or AI
- Human-AI handshake gates progress at human-required steps
Extension of /goal that adds explicit human checkpoints. Prevents Hermes from completing tasks that actually require human action.
Steal forproject management content, delegation frameworks, AI automation content
14:00model
The Pantheon system
Assign specific models to specific missions. Tag the best model for the best job to achieve high performance without burning tokens.
Steal forAI cost optimization content, agent architecture content
How they asked for the click.
VERBAL ASK
23:03next-video
“The most powerful, underrated connection that you could possibly make within Hermes Agent and we are gonna learn that by watching this video right here.”
Ends on a cliffhanger with no resolution. Classic YouTube series hook that forces the next-video click. No newsletter pitch, no sponsor.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
