7 Hermes Desktop Hacks That Will Change Your Life transcript

Hermes Desktop looks like a cleaner way to chat with an AI agent. That is not why I care about it, though. The reason I care about Hermes Desktop is that it finally makes the parts of Hermes that were already powerful actually visible. Your files are visible. Your tools are visible. Your sessions are visible. And your scheduled jobs are visible. And [music] when you set it up properly, it stops feeling like another AI chat app and starts feeling like a command center for your entire computer.

So, in this video, I want to show you seven Hermes Desktop hacks that I have been using every day. These are the workflows that make Hermes feel [music] less like a chatbot and more like an AI employee sitting inside your machine. Let's get into hack number one. The first hack is to stop asking Hermes to immediately do the task and instead ask Hermes to interview you before it even starts. This sounds like a very small thing, but it fixes one of the biggest problems with AI agents.

They start working way too early. Or sometimes they don't understand what we actually want. So, if I say something like, I don't know, find me homes in in France. Hermes can technically start searching. But that is not the real task. The real task is help me find homes in France that fit within my budget, that are within the cities that I would like to be in, that are, I don't know, that have a garden or whatever my preferences are.

That means the agent needs taste, context, budget, location, emotion, what are my deal breakers, what are my big pluses, and the reason behind the purchase. So, instead of giving Hermes a vague task, I use a prompt like this. Help me find homes to buy in France. Use ask user question until you reach clarity. And once you submit that, you're going to see that since this is a desktop app and you're not operating on Telegram, it can do things like embed a questioner.

There you go. Inside of Hermes agent, which is pretty neat way to prompt your Hermes agent. So here, look at this. It's asking, "What's the real buying brief for France? Start with the constraint that changes everything. What is your all-in budget?" Which is super important for a task like this. So it breaks down the task. It thinks, "What are the things I need clarity on? What are the question marks I have? The gaps that I need filled before I run this prompt properly.

So let's say we go, I don't know, 100k all-in. What is the main purpose? The answer decides whether I hunt charming usable rural houses or tourist rental resale safe location. Uh say I want a holiday weekend home. There you go. It's narrowing down exactly what it needs. Uh I want to be able to drive. Shortest drive. I don't know. I'm just answering uh randomly here right now. >> [sighs] >> Lucky for you, I'm not doing uh home shopping uh live during this video.

What renovation tolerance should I enforce? Move-in usable for 100,000 in this economy. Buying a a move-in ready home is going to be tough, but I mean, I It's asking me for my criteria. One bed minimum is fine. Two bedrooms minimum. Two bedrooms minimum. Absolutely can't have any less. What's the minimum for it to feel like a holiday home? A small garden is required. What vibe should I optimize for inside? Northern France, cute village with cafes nearby.

How strict is the cafes nearby? Uh under a 15-minute walk. Final filter, should I prioritize charm, romance, or resale safety when there's a trade-off? Uh always prioritize resale safety. Not financial advice, but hey. And there you go. It now knows my entire brief, and I can tell it go ahead. Now that it knows my brief, go ahead and start with the search. Simple as that. Do you see how with this information Hermes can help you find or do exactly what you want in the exact way you want it?

And sometimes ways you didn't even know, like were I to look for a home, I would not have thought of giving Hermes this descriptor. I want it to be a cute village, but once it asked me, it also allows me to think out loud and to realize my priorities, and it helps it narrow down on exactly what it is you want to find or do. So, if the task has taste, money, risk, personal preference, or just unclear constraints, do not let the agent start immediately.

Make it interview you first. Because a good assistant does not just run with a bad brief. A good assistant asks the right questions until the brief is worth acting on. And once Hermes has asked those questions, the next hack is to not throw that context away. On to hack number two. The second hack is to pin your most used sessions. This is one of those things that sound basic until you actually start using Hermes Desktop for real work.

If every time you open the app, you start a new chat, you're basically just hiring a new assistant every morning and explaining your entire life again. >> I want you to buy me a new home. I want it to be perfect. I want it to be shining. I want it to be in the best location ever. I want it under $5. I want every single thing to be neat and perfect. >> Wait. Who is this? But, that is not how real work happens. In Hermes Desktop, I can keep my important sessions visible and come back to them anytime I want.

So, instead of having to start from scratch every time, I can have one pin session be around my YouTube channel, another be around this home purchase that we just narrowed down on, another for me is investments, another is email. Having Hermes take care and manage my email, flag any missing emails I might have not replied to. Each session becomes a workspace with its own decision trail. That way, when you're coming back tomorrow, you won't have to re-explain the context from scratch every single time.

You can just open up the right pinned conversation. Pin sessions turn Hermes from random chats into recurring workspaces. And once you combine this with the first hack where Hermes interviews you before doing the work, every pin session becomes more valuable over time. And that is a real shift. You're not collecting chats, you're building workspaces that compound. On to hack three. If you use Hermes for real work, you will notice something annoying.

You keep writing the same instructions sometimes over and over again. For me, that might be when I ask it to check posted videos before suggesting YouTube ideas. I tell it go use vidIQ before writing tags. Do not use m dashes. Do not write generic AI guru descriptions. Label evidence properly. Verify the output before telling me it's done. The list goes on and on and on. And there are a set of instructions that I needed to go through whenever I'm uploading a YouTube video, for example.

Or or I can turn that into a skill. A skill is basically just that. It's a reusable operating procedure that Hermes can use every time I ask it something. It tells the agent when to use it, what to check, what workflows to follow, what mistakes to avoid, and what the final output should look like. That sounds boring until you see what it does. Instead of prompting from scratch, you give Hermes the playbook. For example, you could create a skill called filming brief.

That skill could say when I move a YouTube idea on my Notion database from idea to to film status, check whether I already posted it. Validate the current demand for this video. Give me ways to improve the title. List proof assets and write the first 30 seconds. Now, every time you ask for a filming brief, Hermes is not improvising. It is following your actual standard. This is the difference between hiring a smart freelancer once and training an employee to do your work your way and actually making use of Hermes' ability to self-improve.

So, say for example, I go to YouTube and I want I need video idea, for example, you're going to see very quickly that it's going to start using the skills that we've developed together. There you go, skill view. Right away. Didn't even need to think much, just skill view. Oh, we have a skill for that. And instead of me having to explain in detail what it is that I need and how to operate and how to do it and what are each and every single step needed, it will just use a skill right away.

And [snorts] the great thing, the amazing thing and while this does There you go, it's using the vidIQ bridge, it's using every single point in that procedure list after the other. >> [snorts] >> The great thing about Hermes Desktop is you literally have all of your skills listed right here. You can see them, you can turn them on, you can turn them off. You can edit them, you can do whatever you'd like. You can also search up uh the skill, for example, Nova, which is my YouTube uh helping skill, my YouTube agent.

Or I can search up my X agent skill, Sage X. There you go, Charbel's X content strategist. I can turn him off, on, edit him, whatever I'd like to do. Skills are a massive upgrade for your Hermes workflow. But skills are only one layer. The next task is where Hermes starts doing work even when you're not sitting there prompting it. This is one of my favorite Hermes workflows because a normal AI assistant waits for you to ask for it something.

A useful assistant messages you first. That is what cron jobs are for. Cron is just scheduled work. Every morning, every hour, every Monday, whatever you choose. But inside Hermes, Cron becomes much more useful because the scheduled job can actually think, research, summarize, and send you the result. For example, I can have Hermes check for fresh AI news every morning, or scan YouTube topics, or look for X reaction opportunities, or review my content board.

The key is not to make Cron jobs noisy, uh as in don't have a billion different Cron jobs that you never use. A bad Cron jobs, because then trust me, you will just be overloaded with a bunch of information that you don't use, you'll never use, you've never used. I don't know why you even have them. Hermes doesn't even know why you even have them. Why are you even using Hermes? Why am I using Hermes? I don't know anything anymore.

But, let me show you that real quick. In desktop, the Cron layer becomes a lot more manageable because it's a lot more visible. It's right here, Cron jobs. You can expand it, and you will see all of your Cron jobs. So, let's say for example, this Cron job, Notion to film brief watcher. Uh the watcher I was telling you about. If I move an item from idea to to film on Notion, it will uh research that idea. Is it a good idea?

Is it bad? What is a better title? You can either trigger now straight from uh the Hermes desktop app, or you can even manage it. So, if I go to manage, this is what you'll see. You'll see a run history, as well as an edit Cron. You'll be able to change its name, you'll be able to change its prompt, its frequency, where it delivers it to, and whether it has a custom schedule. Here's an example of a custom schedule, only weekdays at 9:00 a.m.

So, it's not every single day. One thing I'd love to see in Hermes in the future is sometimes when I create a cron through Telegram, for example, it won't really show it properly here. For example, this is a cron I've created through Telegram using Hermes on Telegram, and it doesn't show the prompt that I used. It doesn't show the the real frequency uh or the custom schedule that I've inputted. So, I wish it would, instead of, you know, only displaying the ones I have built inside of Hermes desktop.

But, that's it. This is insane because now I'm able to see the things consuming my AI agents' credits and LLM usage and the bill that I get at the end of the month, I'm able to see them and turn them on or off or manage them at a simple click of a button. I don't need to ask, "Hey, list me all the cron jobs you have." I can just go and see and turn them off very easily by literally just pressing pause. Boom. And once you have scheduled intelligence, the next step is event-trigger intelligence.

Let's go. Hack number five. Cron is time-triggered. Webhooks [snorts] are event-triggered. And that difference matters a lot because cron is "Every morning, check what happened." But, a webhook is "When this specific thing happens, wake Hermes up and do the next step." So, let me paint a clearer picture for you. If [snorts] a Notion page moves from idea to the film, like the thing I was telling you about earlier, Hermes can wake up and generate a filming brief.

Say I have a form that I sent to clients. If a client fills out that form and the form is submitted, Hermes can wake up and qualify a lead. If a new competitor uploads a video, Hermes can wake up and inspect whether it is worth reacting to. That is where the system becomes much more interesting. You're not just chatting with an agent. You're connecting Hermes to places where work already happens. You're making it go look at the triggers that trigger you to work.

You're basically telling it, "Hey, instead of me going and, you know, staying online 24/7, you do that. You stay online 24/7 and watch events happen. And if an event happens that interests me, then run that job to it, react to it, create work from it and through it. Okay, I know I'm hammering you with this literally this one example, but I find that's one of the best examples for me to explain this concept to you. It's an example that resonates with a lot of people, which is the Notion example.

So, say I really like an idea, a $0 to $10,000 AI challenge, and I move it to two film {slash} write on Notion, which is where I plan most of my videos, as you can see. >> [snorts] >> What this does is this triggers my cron job, which is linked to a webhook. So, the webhook is always awake or maybe every 10 minutes it reads whether an event has happened. So, say I press trigger now if I want to force it cuz it said in 7 minutes it was going to fire.

If I don't want to wait, I can just trigger it. If I want to wait, then it can just happen in 7 minutes. Once I trigger it or once it triggers itself, then it sends me a message here on whether I should film it or not, what are better title uh ideas for it or even if it's a good idea. This is a great example, by the way, something it sent me right now because I've uploaded a new video, Can ChatGPT actually make Minecraft?

Which is a banger video that you should watch. But anyway, whenever I upload a new video, it sees that and it reacts right away. It starts tracking that. How is that video doing? Uh what is its performance like? Should I change the title? Should I do anything to it? Is its click-through rate high? Is it low? Is its watch time high? Is it low? And starts suggesting edits for me to enhance that video, that already existing upload better.

But there is one more layer that makes this even more powerful. You do not want one agent doing every single job. On to hack six. The sixth hack is to stop using one generic assistant for every single job. For me, the best example I can give you is YouTube. I do not want my YouTube agent to behave like my coding agent or my research agent or my personal administrator. I want a specialist. So I have a Hermes profile called Nova.

Nova is not just a new chat. It is a separate YouTube strategist profile with its own role, memory, skills, sessions, and entirely different workflows. That means when I open Nova, it already knows the job. It knows to check my posted videos before for suggesting ideas. It knows my content board lives in Notion. It knows to use tools like VidIQ before making YouTube recommendations. And it knows how much I care about titles, retention, thumbnails, so on and so forth.

And it also knows what I hate. That is the difference between prompting an assistant and training a specialist specifically. And the lovely thing on Hermes agent desktop is you can switch between them so easily. I can go to Nova, for example, and load up my Nova specific sessions my PIN chat. I don't have any, but uh if you did have a PIN chat with that specific agent, then you could access that that quickly, and [snorts] you could even manage their profiles.

For my default, this is my profile. For Nova in specific, she has a different soul bot and D file, uh which all the other ones share. But, if you'd like the soul of your agent to be different or any of your specialist to be different, then this is where you can customize it. You could rename them. You could even change the model that that specific agent uses. So, if I don't know, uh you need an agent to be on Anthropic's Claude, for example, you can just set it up so that every time you open that agent, they are already using the exact soul file you want, the exact skills that you want, the exact model, AI model that you want them to have access to.

And that's one of the best things. You can also create new profiles straight on Hermes desktop. One of the best examples I like to give here is you do not want your dentist to be your heart surgeon. Uh I mean, both of them are great at their job, but I mean, I don't really need to explain why your dentist shouldn't be your heart surgeon. Profiles let you separate those jobs instead of forcing one messy assistant to do all of the different jobs.

And once you see it in this way, Hermes desktop stops being one AI chat app, and it becomes a place where you can manage a team of specialist agents. But, that being said, on to the last hack. The last hack in this video is one that I don't think many people will especially go crazy for, which is why I've included it at number seven, but it's one that I think more people are going to care about over time. >> [snorts] >> Hermes desktop does not have to only talk to the Hermes instance running on the same computer.

You can point it at a Hermes back end running somewhere else, whether it's a VPS or a home server or a Mac mini, a separate device, a machine that stays on 24/7. So, for me, I have a Mac mini running 24/7. It is where my main Hermes agent instance lives. Yet, the PC that I use the most, the setup that I use the most is my laptop. So, I don't want to have to switch between both devices every single time, which is great because I can connect Hermes desktop to the same Hermes instance as the one on my Mac mini, as the one on my local device.

And if you'd like to do that yourself, all you have to do is click on this setting icon and go to gateway. And this is where you can choose what gateway you want, whether it's local, so one that's running on your current machine, the one that you're operating right now, or another gateway, which is what I have picked. You input the gateway's URL. You can ask Hermes itself to help you set it up. You don't have to worry about all these technicalities.

You can test remote. Once it's tested and connected and it works, save and reconnect. And that's literally it. You can access the logs here in case there's any errors. You can troubleshoot with that. The great thing as well, and that's maybe an advanced hack, inside this hack, is you can connect different agents to different gateway instances. How cool is that? You can I don't know, if you have multiple devices or if you have one instance, one specialist agent on one device and another on another, then you see where this is going.

You can connect them to different instances. Or if you have one on VPS and one on your local device, then you can connect both all under one uh surface, all on the same application, which I love. I love being able to access all of my Hermes instances in one simple UI. One simple application. Look, Hermes Desktop is not interesting because it makes your Hermes agent prettier or the chat interface prettier. It is [snorts] interesting because it makes agent infrastructure easily understandable and easy to access with project folders, sessions, skills, cron jobs, webhooks.

I mean, we've covered a lot in this video. Those are the pieces that turn Hermes from something you prompt into something you operate. If you only use Desktop as a chat window, it's nice, don't get me wrong, but if you use it as a control room for your files, your workflows, your jobs, your specialist agents, it becomes a completely different thing that you're using. And if you connect it to a machine that stays on all the time like a Mac mini or a VPS or just any machine, that is when it starts to feel like a real AI employee.

Not because it's magic, because it finally has somewhere to live. If you want me to make a full walk-through of my exact Hermes setup, how to go from zero to 100, then comment the word setup and I will turn the whole thing into a step-by-step guide for you. And if you've enjoyed this video, make sure to subscribe. I have a ton more just like this and oh, would you look Look that? The YouTube algorithm gods think you will really enjoy this video next.

So, I'll see you there.