NEW Hermes AI Browser Agent: Automate ANYTHING?

追加: 2026-05-08

[00:00:00]Hermes' agent just got a brand-new free skill, and it lets your AI agent completely take over your browser. We're talking clicking, searching, booking, filling out forms, finding leads, all of it while you do something else. And the best part is it gets smarter every single time it runs. It actually learns from what it does and gets faster each time. In this video, I'm going to show you exactly how to add this skill to your Hermes agent in just a few minutes for free. You don't need to do any coding, no technical experience is required, nothing like that. I'll show you how to set up, give it tasks, and how to run it in the cloud so that it's working for you around the clock 24/7 even when you're not at your computer.

[00:00:43]By the end of this, your Hermes agent will be able to do browser automation work that you used to spend hours on.

[00:00:51]Let's get into it. There's a brand-new AI skill for Hermes agent that allows it to control the browser. It's a browser use agent, right? So, this is with a new skill called browser harness by browser use. And literally what you do is you take the GitHub link, which we've got right here, all the details of this, and we're going to take that, and then we're going to send it over to Hermes agent and teach it a new skill so that it can control our browser for us. So, let's get straight into this. So, I'm going to say, "Okay, let's copy this." And then we're going to paste that in here. All right. So, you can see here there's a ready-to-use setup where you just copy and paste the prompt for LLMs, and then you can give that to your Hermes agent so that it can then start controlling your browser for you, right? Now, browser harness is basically a self-healing harness for AI agents, right? Which lets AI agents edit their own helpers complete any task. So, you can run your agent locally on your browser or 24/7 inside browser use box as well. This is kind of like a sandboxed area where AI agents can just go off in the cloud as an AI agent and start working. So, you can see here Hermes is now beginning to do this and set this up. And if you're wondering, "Okay, what does this do? How does it work?" etc. So, basically you can connect an LLM directly to your real browser with a CDP harness, right? This is for browser tasks where you want your agent to control it. So, you can basically teach Hermes agent to just go off, browse the web, and connect to Chrome, right? It has its own agent workspace. And this is a free browser use agent, right? This is a free skill you can teach it. And the good thing about this is you don't really need to be technical to set it up. You've just got Hermes agent running, you give it the skill, and it goes off and installs it. Now, you might be wondering, "Okay, what's new about this?" cuz there are AI agents before that have been able to use browser use. Well, basically there's three new super powers to this, right?

[00:02:42]Number one is self-improving browser tool. So, the agent doesn't just do the tasks, it also learns how to do them better each time. So, when it figures out a clever way to do something on a website, it saves that knowledge, and then next time it does that, it already knows exactly how to do it, right? So, literally rewrites and edits its own helper mid-task. Right, the harness improves itself. It gets better and better browser use the more you use it.

[00:03:05]You can also run multiple browser sessions at the same time in the cloud, right? So, these are browsers, meaning websites, can be used by AI agents. And you can see here it's come up with a pop-up on how to use this. So, we're going to click allow. And also it can rewrite code to improve itself and get better. So, examples of how you could use this. So, for example, like for lead generation and research online. It could be, for example, even shopping online, filling out forms, researching the web, booking appointments, stuff like that.

[00:03:33]Like super powerful for this sort of stuff. And you can see Hermes is now testing it. So, it's actually asking like, "What's visible on this page?" and trying to understand what's going on.

[00:03:40]So, you might be wondering as well like, "How does this work?" Well, basically you type, for example, like I don't know, "Go find me 50 agency leads."

[00:03:47]Hermes agent receives the task, browser harness opens the browser, the agent navigates to relevant websites, searches, scrolls, collects information, and researches. If it hits a problem, it writes new code fix itself, and then the task is complete, and you get the answers back inside your terminal here.

[00:04:04]So, the agent figures out every step by itself. And you just want to make sure you have this set up inside Chrome. This is a new update that came from Google where they allow agents to connect to the browser, and you can see that this is now connected to Hermes directly, which is pretty cool. And so, how does it work step by step? Well, you install the browser harness, which we've done today with this prompt. Then you're going to enable remote debugging in Chrome, which we've done over here. Make sure you've got that checkbox connected.

[00:04:32]And then if you want to, this is optional, but you can connect to cloud as well. So, you can go to cloud.browserless, grab an API key, add that as a environment API key inside your browser harness folder. And basically what that will do is allow your agent to use a cloud-based environment for browser use agents in the cloud, right? So, it's basically got this browser, this personal computer in the cloud that it can access as well.

[00:04:55]And then you can just give it tasks, you know, some examples like this. So, you can see here, for example, it's just navigated to GitHub just to open up and test, right? So, it says new tab, and it's connected directly, and it can go directly to the browser now and control it. You can also, after each task, check this folder. And by the way, I think it's scrolling now through the website.

[00:05:17]So, you can see here it's using vision analysis to check the page, have a look, it's capturing screenshots, etc., and it can navigate the browser. So, after each task, you can check this file to see what it's learned as well. So, it creates a agent helpers file where it writes new helpers for future tasks, and it just makes future tasks faster and faster. And then, as an example here, it said, "As a quick demo of how the interaction works." This is what Hermes agent sent to me. "Should I click on the start button for you?" And then we can just type yes, which is right here. So, let's click yes here. See how it does.

[00:05:51]And you can actually see what it's thinking here. So, it's using vision analysis, and it's like, "Okay, here's the coordinates of the page. I need to click on it." And by the way, I don't think this is as good as, of course, just navigating the computer or the browser yourself, but this would be useful, for example, in situations where the agent needs to do something like this. And also in situations where you give it a task, you go off and do something else, and then you come back later, you know, like schedule tasks and that sort of thing. Also, it's going to depend on how good the model is. So, for example, if you're using something like Claude, Claude is an API, is one of the best for computer use agents. Now, if you also want to get the computer in the cloud and create like a 24/7 cloud agent, you can do that at cloud.browserless, and you can test it for free. You can also set that up with Telegram and integrate it with, for example, like Discord or Dropbox or whatever you want, and it will take about a minute to set up, and then from there you're good to go. Now, you can also see there's a green button next to this tab, which means that the agent is controlling it. And then later what you can do is actually track all your agent sessions, your remote browsers, etc. So, this is something that we actually tested a week ago, and we can see our previous sessions right here. We can also grab an API key inside our settings, and then connect that to Hermes with our 24/7 agent box. And just to recap on this, the browser harness is a new skill added to Hermes agent. It connects your AI directly to your browser use. The agent can browse the web, click things, research, etc. It's also self-healing, so it fixes its own mistakes and gets faster every run. You can run stealth cloud browsers in parallel and have three for free with the browser use cloud setup. And you can also give it pre-built knowledge with domain skills, which is pretty cool, so that it understands how to use popular websites. You can also set this up inside Claude code or open Claude. It doesn't have to be Hermes. And if you want a full step-by-step guide, tutorial, 30-day road map, etc., and example prompts, you can get that full guide inside the AI Profit Boardroom.

[00:07:50]This comes with new daily advanced tutorials, like you can see. This is my AI automation community that helps you save time, grow, and scale with AI automation and AI agents. Link in the description or go to the AI Profit Boardroom.com to check it out. We have 3,000 members inside here, tons of trainings and useful tutorials, four weekly coaching calls where you can meet people who are also using AI agents, share your screen, ask questions about your setup. And then you can also meet people in your local city who are using AI agents just like you. Plus, you can connect with me personally inside here, too.