Hermes Agent: NEW Goals Update is Insane (FREE)!

Hermes just dropped a free update called goals and it makes your agents way more powerful and autonomous. You type one goal, Hermes chases it step by step all the way to the finish line without you having to prompt it every five minutes. That could be for content, research, lead generation, full websites, one command, and it keeps going until the job is done. And there's something in here I'm going to show you that most people are completely sleeping on, and it changes what you can actually get done with Hermes every single day.

This is a new free skill inside Hermes, takes two minutes to get, and by the end of this video, your Hermes agent is going to be running full multi-step projects on its own. Let's get into it. Hermes has a brand new update called persistent goals, and this allows Hermes to run autonomously as an AI agent using the goal feature. So you can see here, this is an announcement from one of the co-founders of Hermes, and he has said that this feature was inspired by Ralph Loops.

And what this means is if you use goal, it will start a loop with a supervisor model determining whether the task completed at the end of the agent loop. If it hasn't, it will continue to go until it's actually completed, right? So this is a brand new feature inside Hermes agent. You can get it for free by just updating inside your terminal.

If you run Hermes update inside your terminal, you can pull in the latest updates from Hermes agent. And then from here, we can start getting this set up. So the way this works is this has four different commands, right? So you can run goal with your text, so you set standing goal.

So you can say, okay, goal, build out an SEO website or an AI automation, et cetera. From here, you can set and see the latest status. So if you put forward slash goal status, then you can pause it, you can resume it later, and you can clear it as well, right? If you just want to end the goal.

And this works on CLI and also the gateway via the central registry too. This is kind of like an autonomous version of Ralph Loops. It was very popular with Claude Code. Basically meaning the AI agent has a goal and just keeps going at it until it's completed.

So we've updated now, as you can see, and then we can put forward slash goal and the text. So if we type forward slash goal, you can see this command right here. Just make sure you scroll down and select it. And then we need to set up a goal for this agent, right?

So if we say, okay, build a new blog about Hermes AI agent, optimize it for SEO, plus continue posting daily until it's completed and ranking, you can see that this is now initializing the agent with the current setup, right? Now, what you can do once this is completed or whilst this is running is you can do forward slash goal status, and that will actually show you the current setup. Now, what you can also see here is that we have a turn budget. What that essentially means is that it's got 20 turns to get this job done.

You can change that if you want to. And so this is a goal, and this is how many turns it has to complete that task, right? And you can see here after each turn, a judge model will check if the goal is done. Hermes keeps working until it is, or until you pause and clear it, or the budget for turns is exhausted.

Now, you can use goal status, goal pause, goal resume, and goal clear to manage that goal. So if we type in goal status now, and by the way, if we scroll down here, you can see it's beginning to actually deploy the website to Netlify as well, which is pretty cool. And we have a plan with the number of tasks being completed too. Now, you actually see that there's a goal loop, right?

So you can see this looping over several times to check the status of the goal and see where it's at. So here it said, I'm building an SEO optimized blog for Hermes agent. I'll create a dark themed fast static website with meta tags, schema, markup, sitemap, and semantic HTML, then deploy it to Netlify. Now you might be wondering, okay, how do you use this?

Well, for example, it could be like, okay, fix every error in this particular topic. Port feature X from repo Y, et cetera. Investigate why this isn't working. Build a small CLI to rename files.

And then after this, you see the goal accepted, which we've got here. Then turn one is going to run just like a normal message, which we've got. After that, the judge model decides if it's done or whether to continue and the loop continues if required. And then eventually, once that's completed, you'll see goal achieved or goal paused.

Now, if you want to change the continuation turns, aka how many times it runs, you can change that in the goal.max turns figure in config, as you can see. And also if you send any messages whilst the goal is active, that will take priority over the continuation loop. And what's interesting here is you can actually set a goal, close your laptop, come back tomorrow, and then you can run forward slash resume, and the goal will continue just as you left it, right? So if you close the terminal here and the goal stops running, that's okay.

You can resume it later tomorrow, right? And you can see how many minutes it's been running for here. This is how to change the setup inside the config for goals in terms of the max continuation turns it can run. And then you can also change the judge model, right?

So the AI is judging whether the goal is actually completed, and it uses the goal judge auxiliary task, right? So you can actually change the provider and the model that you use to test if the job is done. Bear in mind, this is a relatively new feature, so just be aware of that. But it looks really cool and something super powerful to test out, especially if you have a big project, and you just like go off and do that, you know.

If you want to clear it, you can just run this in your terminal, goal clear. And now you can see it's actually beginning to write the HTML files for the website too. So this is the website that's beginning to build out with the articles, etc. With the articles as well, so you can see.

And obviously that will take a while to load, but that's basically how it works. So just to recap here, what is goal? It's a Hermes command that gives your AI agent a persistent objective it chases until completion. Why does it matter?

Well, it eliminates the need to keep prompting. It's just one command that creates a full autonomous work session. How does it work? After every turn, a judge AI checks if the goal is done.

If not, Hermes continues automatically. And it has a budget so that it doesn't go out of control, right? It has a term-based budget. If you're wondering, okay, what makes a goal prompt good?

Well, specific, clear, defined, and measurable works. And then what would you actually use it for? Well, you know, content, production, research reports, code fixes, anything that's multi-step, right? And you can use these key commands to handle it.

Now I've actually got a 30-day roadmap and some example prompts and use cases of how you can use it as well. That's inside the AI Profit Boardroom. And you can see here that it's beginning to build out all the blog posts as well, which is pretty crazy. So it's creating a full website just using this autonomous process.

So that is how to use Hermes agent goals. If you want a full guide on how to do that, you can get that inside the AI Profit Boardroom. This is my AI automation community where you can ask questions. You can jump on weekly coaching calls.

You can meet people in your local city who are using AI agents like OpenClaw and Hermes and AI Automations. And also inside the classroom, you get all of my new daily tutorials, step-by-step guides, and you can connect with me personally too. Now, if you go to this section, you'll see the full guide on Hermes goals and how to use them, et cetera, how to get the most out of them. And that's all inside the AI Profit Boardroom.

Link in the comments description or go to the AIProfitBoardroom.com to get access. Thanks for watching.