New Hermes Subagents Update is INSANE!

So Hermes agent now supports asynchronous sub-agents. To get this, this means that basically Hermes now has a delegate tool. And so you can use that to spawn sub-agents. And I'll show you some examples for us how to use it today.

First thing you want to do is just make sure you update Hermes. So you can go inside your terminal and then just type Hermes update, or you can go inside your dashboard for Hermes and just click the update button. And then from there, what it's going to do is basically when you've got a task and you say, okay, create some sub-agents for this task, it's going to use a new tool called the delegate tool. And that can create multiple different agents at once.

So let me show you an example of how that would look. So you could say, okay, run a SEO funnel for how to build an AI agent. And then you've got multiple workers working together, like you can see here, and those workers work to create something awesome. And I'll show you an example of where we've used that for SEO actually in a second.

And the great thing about this is like the old way is it would freeze your chat. The new way is like, this is running in the background and then you dispatch the workers and they just run, right? So the old like delegate call would block your chat until the task was finished. With background equals true, this hands control back to you in two milliseconds.

So the work runs for you in the background. So here's an example. We could give our Hermes agent one keyword and then we could have a hive of AI agents working in parallel on five different articles, for example. Then we could have a deploy worker that actually deploys the content and we get that live to our website.

So you can see an example here of how we've deployed just as a total example. We've got five different agents that work to sub agents to create five different articles on our website and deploy them. And so what this means is like, number one, you can use the chat whilst you've got agents working in the background with Hermes. And then number two, these can actually complete real world tasks in the background, like actually creating blog posts, something a lot of business owners use.

And you can see that was all fully done and automated using this sub agent process. So it's pretty powerful, quite easy to set up. And this is just something that got released this morning. We've already set up today.

So I know if we can do it, you can do it too. And you can even get, for example, Hermes to set up for you, or you can get Claude to plug in. What we've actually done is inside our agent operating system, we have multiple different agent profiles. And then if we give it a task for sub agents, it can just go off and build those in the background, which is great as well.

And again, if you want to manage an update to the latest version, you would just update over here. And then if you want to manage the profiles for different agents, like we've set up, you can manage that over here. But basically these sub agents are like kind of temporary agents that work at one time. So this is how it works.

Let's say, for example, you've got a lead Hermes agent here, which is the main chat, the main profile you're using. That's going to send out workers and they all go off and do stuff at the same time. And that's how we created that example. Then it's going to give you the results and that goes back into your chat.

So you can have a lead agent, you hand it one goal, for example, like run the funnel for this keyword, and it decides the fan out. So it never does the work itself. So as an example of this, you can see these agents are just working in the background with five different workers when we give it a task. So it's using delegate task, background true, and these will just get dispatched in the background.

And then you've got the workers. So you've got one background sub agent per job, each dispatched in two milliseconds using the delegate task tool, and each runs its own thread with its own goal and context. They all work at the same time, which is great. And they're off to the side of your chat.

Kind of like how, if you ever use Cloud Code, this is how it works as well, it's a similar way. And that means you can keep planning, you can keep chatting with your agent and start the next keyword. And so as each worker finishes, it results back and re-enters your chat as a new turn, self-contained, and then this can all live inside your Hermes agent. So this can all live inside a Hermes chat here.

We can give it a task. So for example, we could say, okay, you know, delegate a keyword research task to five sub agents working in the background, and Hermes can just go off and implement that for us. Now, also the other thing to note here is they don't finish together. So each result flies back in the moment it's done, re-entering your chat.

So you could have like one worker that finishes first, then another worker finishes later or two finish at the same time, et cetera. And so if you look at like the old way of doing this, you know, you've got one at a time, right? Each article would wait for the last, and that would be a very long warning. Whereas with this, you've got a hive and all five start at the same time, same five articles, they all queue together, and they're just running in the background.

That's how it can work. So like the old way would be like writing one article, wait for the next one, checking it, each article blocks the next one or each task blocks the next one. You have to sort of sit and watch as the chat is frozen. Whereas the new way is like you've got one command, you've got five workers working at once, they all write the articles, you walk away, everything comes back on its own, and the result is a whole funnel that gets built whilst you're doing something else.

So the cool thing about this is like, before people were like, well, if I'm not watching each step, it will go off the rails, et cetera. So each worker gets one tight goal in its own context, and every result flies back for you to check, right? You're basically the lead, you set the jobs, and then you walk away. Also, other people say, well, running agents in parallel sounds like chaos, but it rides the same rails as like a background terminal job.

Other people say, well, this is only for coders, but you can speak to it in plain English, and then it will just go off and build stuff. So just to recap, you learned how to stop waiting and watching, dispatch is two milliseconds, so the work just runs whilst you keep going, you stop babysitting the funnel, you shipped everything in parallel, you collect, you don't have to chase things up, so it comes back to you inside the chat, and you don't have to just do this for SEO, this is just an example, and it's one command. So Hermes update, and then give it a goal, and it would delegate stuff to you. It's kind of similar to, if, for example, you ever seen UltraCode with Claude, where it has the main orchestrator, and then you have agents working on different tasks based on that one big goal, basically you're just delegating everything to your Hermes agents in the background.

But also, the cool thing about this is you could have like, for example, an agent with a different API working in the background, and if it's just like a menial task, or an admin task that's working for one sub-agent, then you could use a cheaper API, or you could use a free one, you could use a local one, et cetera, and it would still work. So that's basically it. Now, if you want your agents to multiply your output, not just be one more tool, grab the agent operating system inside the AirProfit One, which turns Hermes, Grok, Build, Claude, GLM, and Kimmy into one system with shared memory and one dashboard. Every new feature, for example, like asynchronous sub-agents, makes the whole thing more powerful automatically.

And inside the AirProfit Boarding, link in the comments description, go to the airprofitone.com, you get the full agent operating system. You get the Obsidian memory setup, a coaching course where we build life together, a 30-day roadmap, daily tutorials, a member map, and 24-7 community. And, you know, you might think, well, this is awesome for you, Julian, but, you know, what about if you don't have time? Well, we've actually had so many people building out their own agent operating systems using our setup that I know that if it works for them, and if it works for me and none of us are coders, then we can all get great results with this.

So it's definitely worth checking out. So I hope to see you inside there. I also answer questions inside the community. You can ask questions yourself as well inside here.

Inside the classroom, you get access to all of our best trainings and the new daily agent operating system. Inside the calendar, you can actually jump on weekly coaching calls. You additionally get a map so you can meet people in your local area using AI Agents like you. And this is all inside the AirProfit Boarding.

Thanks for watching.