Hermes Agent Goals is INSANE (FREE Setup)

Hermes Agent's new goals update just dropped and it turns Hermes into a true 24-7 AI employee for free. Imagine telling your AI agent exactly what you want done, walking away and coming back to find it completely finished. That's exactly what Hermes Agent's new goals update just made possible and is free, which means that you don't need to manage it. You don't need to type keep going over and over. There's no being stuck as the bottleneck yourself. Just give Hermes a goal and it runs on its own until the work is done. This is one of the biggest free upgrades Hermes has ever shipped and in this video I'm going to show you how to use it to get more done in a day that most people get done in a week. Plus there's one thing I'm going to show you later that makes this whole system even more powerful. Stay with me for that one, let's go. Today I'm going to be walking you through the new system inside Hermes for goals. Basically what this means is you can set a goal for Hermes AI Agent. So for example, we can go inside here and we can say, okay, go off, create a blog about OpenClaw and make sure it's hosted. So I can preview it, add a few blog posts, plus make it look beautiful. So we can plug that into our system. Then we can hit launch goal. And you can see that's now running. And so what this is going to do is move into goals mode. And basically what happens with Hermes Agent goals mode is you set the target and then you can just walk away. Right. So Hermes forward slash goal gives the agent a standard objective that survives every turn. Right. So a judge model actually decides if it's done or continued after each response. And if the goal isn't met, Hermes feeds itself the next prompt and keeps working until the target is hit or the turn budget actually runs out. So you can close your laptop and then you just come back to finish work. Right. So this is Hermes Agent and that's the goal over here. Pretty amazing stuff. And the default for this is 20 turns, which means that it would just keep going off and making sure it's done 20 times before it's finally completed. This is the power of this stuff. Right. So you can set the target, walk away, hand Hermes a long horizon goal. Then it will run that. You can close your laptop, go to sleep, come back to finish work. Right. That's how powerful this stuff is. And so you might be wondering, okay, what is Hermes goal mode? For example, if we look at this system, you see how it's running this autonomous loop right here. To get the goal done. That's exactly what this is about. Right. So forward slash goal gives Hermes a standing objective that doesn't stop after one term. After every turn, a lightweight judge model reads the agent's last response and decides, is it done or should we keep continuing? If continue, Hermes actually auto creates a continuation prompt back into the same session and keeps working without you typing, keep going three more times. Right. So, for example, if you go into ChatGPT, you say, go off and do this. It would just stop on one turn. If you go into Hermes agent using the goals feature, you can see how it's still running autonomously until eventually gets done. That's the difference. Right. And it's Noose Research's take on Ralph Loop. Now, if you're not familiar with Noose Research, basically the creator of Hermes agent, one of the most respected AI labs in the world. And this is their take on something called the Ralph Loop. Right. So you keep a goal live and it doesn't stop until it's achieved. This is something similar to what Codex just released, which is forward slash goals. Well, but adapted to Hermes architecture. Right. And so it works identically on the CLI and every gateway as well. So you can do this inside Telegram or Discord or inside your agent operating system like I am. And Hermes will basically go off and do this. Right. So the old way and the new way. Let's talk about this. So let's say, for example, we had a same long horizon task. Right. Same agent. So you're using, for example, Hermes, but two completely different workflows. Right. So you've got the goals workflow and then you have the old way, which is like after 30 minutes of keep going, get it done. Right. With the old way, you'd have to reprompt it every single turn. Right. And it'd be like 30 minutes of just keep going. So you type the goal. The agent does one turn and it stops. Right. Then you type keep going. The agent does one more turn and stops. You type again. Right. And you repeat that 10 or 20 different times. And each prompt loses more context because the agent forgets the original goal halfway through. And if you walk away for an hour, the agent is sitting idle, waiting for your next keep going. So you have these long refactors that take hours of human in the loop attention. And the goals get abandoned eventually because nobody wants to sit and just babysit a Hermes agent. Right. They just want to let it go off and do things. And so agents with the old way basically do small tasks. And you're basically the bottleneck. Right. On anything larger because you have to sit there and click continue. Keep going. The new way is Hermes forward slash goal, which means zero minutes of just attending and making sure it gets stuff done. Right. So one forward slash goal with the goal inserted and the agent starts immediately. Then the judge model decides if it's done or continues after every turn. So there's no human needed. It's a 20 turn budget by default, which is configurable. So you can actually change it anytime you want. And it can auto pause at the limit. So you can resume in chunks. The original goal text stays in context every turn so that the agent never forgets what your original goal was. And it persists across forward slash resume. So if you close the laptop, come back tomorrow, the goal is still standing. Right. And you can also set up a sub goal with the text as well. And this actually tightens the success criteria mid loop without breaking everything. So you set the target, you walk away, you come back to finish work. Right. And so with the old way, the result is like the agents do small tasks, but you don't create anything bigger. With the new way, you can have long horizon work that finishes on its own and you stop being the bottom. You don't manage the agent turn by turn. You actually just hand it the target. That's the difference here. And you can see it's still running right here. Right. It's very autonomous. Let's see what questions we've got here. Sportus says, side question. Do you think the new Google updates are impressive? Which LMS do you prefer? So one of the coolest ones I've discovered recently for LMS is Grok 4.3, particularly if you're using Hermes agent. Very powerful stuff. With the new Google updates, the one that I love, and I think people are missing out on this, is anti-gravity. Right. Because you can use anti-gravity CLI inside the system like this. And then you can have it working across all of your AI agents in a beautiful dashboard. That's what you want. Right. You can see anti-gravity over here. That's what I like. And I think you want upgrades and stuff. So for example, you've got like Omi, you've got the video generation, stuff like that. That's really cool as well. And also just in general, if you're using Google, it's way faster now on their AI side. So I like that too. So let's talk about the three moving parts here as well. So some people say, what is a judge about? After every turn, Hermes basically calls the judge with standing goal text, the most recent response. And it's deliberately conservative, right? So it only marks a goal done when the response explicitly confirms completion, when the deliverable is clearly produced, or when the goal is genuinely blocked. Right. And so it helps you achieve progress. Then you've got the turn budget. So you can say the default is 20 turns, 20 chances to hit this. Right. But it will auto-pause and you can always extend the number of turns that it actually gets to keep going at the goal. And then you have the loop, right? The continuation prompt, which is a plain user role message appended to history. Right. So basically what this means is like every time it keeps running the loop, it's still got the context. That's what we want. You've got all these different commands that you can use with it. So you've got like goal, you've got forward slash goal status, goal pause, goal resume, goal clear. These are all different ways that you can command the agent to, for example, see the status, what you're doing, or see how many turns it's used and that sort of thing, or what agents and what sub goals it has as well. Right. So you have one goal in many turns out that's a play right here. Right. So if we go back, you can see it's still running, right? It's still building out this system right here for us. And then we can see what it's created and everything else. Pretty nice. And then if you want to see like a previous example, you can see that it created out this whole website. Like you can see right here, we've actually completed like different goals. This was just a very simple one. I actually did like a nice little haiku about a lobster, but yeah, you see how it works. So it's really cool. And then you could use this for creating new websites, for example, automating big goals that you have and everything else. And also something to note here is persistence, right? So the goal state lives locally. If you come back tomorrow, if you close your laptop and want to come back tomorrow, then you can run forward slash resume and the goal stands where you left it, which is pretty cool as well. So really cool system. Some people are not going to use it, but the setup is free, right? Because Hermes Agent is free. It's open source. The API that you can use with it, for example, like our alpha, which is free on Open Router is free. And the goal skill as well is free, right? It's created by Hermes and baked into Hermes directly as well. And I've seen so many people get great results with stuff like this, like Hermes Agent particularly. There's nothing stopping you from doing it as well. Now let's talk about some limiting beliefs that might be holding you back here. So some people say, I need to watch every single step. But every goal runs in its own directory. So nothing actually collides with your work. So it's all siloed off, which is awesome. Other people say always the judge isn't good enough and that sort of thing. It might get it wrong sometimes, but 90% of the time you're all good. You're not going to have a problem. And then other people say, I'll just keep typing, keep going to get the same result, right? But if you do that, if you just use AI agents normally, then you are the bottleneck. Because as soon as you walk away, the agent stops working. Whereas if you use forward slash goal, you set the target, you walk away and you come back to finish work. I've actually got a 30-day roadmap inside the AI profitable and for using this. So Hermes goal mode, just to recap, there's three moving parts. The judge model, the term budget and the continuation loop. And the result is long horizon work that finishes on its own. So you stop being the bottleneck on tasks, bigger than one turn. You set the target, you walk away, you come back to finish work, right? Now inside the AI profit boardroom, if you want to get my agent operating system, the goal mode configured inside there. If you want weekly coaching calls in the rest of the agent operating system, plus a 30-day roadmap for implementing this alongside Claude, OpenClaw, Codex, Anti-Gravity, Gemini, Free Code, all wired together inside this beautiful system, as you can see. Then you can get that inside the AI profit boardroom. Link in the comments description or go to the AIprofitboardroom.com. You get all of my best trainings inside here. We've got a zip file for installing this agent operating system. We have all the prompts and also show you when I've last updated it as well. So you know you're getting the current version. And then also inside here, it's just an awesome community of good vibes, people helping each other, people asking questions, people sharing cool stuff, which is great. Lots of people getting awesome wins with AI. You can connect with me personally inside there. You get all of my new trainings. You can jump on weekly coaching calls. We have four weekly coaching calls. And then also you can connect with people in your local city who are using AI agents just like you, right? So it's a great way to connect with people and connect with me personally as well inside there. So feel free to get that. Link in the comments description or just go to the AIprofitboardroom.com to get access. Thanks for watching.