This episode introduces Headroom, a free open-source project trending on GitHub that compresses what AI agents read to reduce token usage by 60–95% without losing meaning. It’s described as a “zip file” for agent context—files, tools, search results, logs, and conversation history—helping agents run faster, cost less (especially via APIs), and forget less due to context-window limits. The script claims Headroom can plug into many agents and tools (including Claude Code, Codex, Cursor, OpenClaude, and Hermes) and offers proof examples such as 92% token savings and compressing 10,144 words to 1,260 while finding the same log error. It outlines a three-step approach: crush, keep (shared reversible memory), and compound (learn from failures).
00:00 Token Saving Breakthrough
01:07 Why Agents Burn Tokens
03:00 Headroom Zip Compression
03:03 Quick Install Demo
03:27 Three Key Benefits
03:39 Goldie Framework Steps
04:31 Proof and Benchmarks
05:31 Common Objections Answered
06:03 Recap and Next Steps
06:45 Boardroom Offer and Outro
Full transcript
Hermes agent have just dropped a brand new web UI for their agent. And this just makes it 10 times easier to manage everything in one single place. You can see, for example, you can switch between sessions model. And they also added a new section here called chat, where you can chat directly with your AI agent inside the UI here, which is really cool.
You can also see the model you're using. You can manage everything from one place and we've already plugged it into the agent operating system. So this is the announcement. The Hermes web dashboard just got a major overhaul.
It's now a feature complete admin panel that manages everything entirely from your browser. So what this means essentially is like, instead of having to go to the terminal or to the Hermes desktop app, you can see everything in one place, makes it easier to manage, saves time and saves a lot of energy going through everything. The other cool thing as well is you may have a mission control where you chat with your agent normally, you've got your whole workspace, but then also how do you manage all your models and your sessions, how do you see everything across all different platforms? You can do that using this.
So you can see, for example, you can see our recent sessions. We can see the models that we've got plugged in right here. And we can actually switch between them, which is pretty cool as well. So we can actually select, okay, we want to use this as the main model or auxiliary tasks.
Now, what that means essentially is like, you can use this for like sub-agents and then you can have this as a main agent model right here. And it's super easy to like switch between your different AI agents and different models here as well. So we can switch between the models inside the list and then we can configure an agent for our sub-agents as well. So what that means essentially is like, it's really easy to switch between models and no longer do we need to go into terminal and then start typing Hermes model and switching between everything, which is quite time consuming and not fun at all for most people.
And also the other thing I would say here is like, this is now way better for being adopted across people who are non-techie, right? So for example, previously AI agents were limited to people who were willing to use a terminal or willing to set this up locally with cloud code or something like this, whereas now you can use Hermes agent directly inside a browser and you don't need to have to use terminal to switch between things. So it's a really cool idea. You can see the new details right here on the web dashboard.
I'm going to break it down for you and explain exactly what's different. And I call this the Goldie command deck. So you can run your whole AI from one single screen. And until now, really setting up your AI agent meant editing different technical files, typing terminal commands, but Hermes have just released a clean browser dashboard that turns all of it into the point and click option.
We've already wired into the agent operating system as well. So the cool thing about this is if you're running an agent operating system, you can actually give your setup plus the documentation for web dashboard, and then say, embed this inside my agent operating system, like you can see right here. And then that's great because it gives you another option to manage your AI agents. And it makes everything easier to control, particularly for switching models and chatting.
So this is really cool. And also you can see here, you can manage all of your channels in one place. So if you want to configure Telegram inside the agent operating system, before we had to do that manually inside the terminal, now we can just configure it with one click and we can just put in the details right here. So that's great as well, because then we can configure everything in one place.
We can change the channels that we use it. We can organize this way better and way easier and it's faster now as well. So this is really cool. And the thing that I would say is you shouldn't need to be a coder to run your own AI agent.
The new dashboard puts every control in front of you. And the agent operating system puts that whole command deck on one screen. Now, why did running an AI before feel scary? It was intimidating, right?
Especially if you've never used AI before, because Hermes is so powerful, but until now, managing it was like fixing a car by climbing under the hood with a wrench. So if you wanted to change a setting, you would open a settings file or your terminal full of code and hope you don't break it. If you wanted to add a key or schedule a task or check what it did, we have to type commands into a terminal window. For a normal business owner, that's a wall.
And it's the reason most people never get past hello or just use it for a day and then stop using it. The AI was always ready, but the wrench and terminal setup was the thing that was scaring most people off. So with this new dashboard, in plain words, Hermes have just shipped a new dashboard, which is a clean control panel that opens in your browser. Now I really like this as well, because then I can manage all my agents in one place, so I don't have to switch between tabs or apps or terminals.
I've got everything in one place and it's just getting easier and easier to manage everything in one place, which is great as well. You can also see it's responsive on mobile, which is great too. So if you're using a mobile device, you can use it like that. And like I mentioned before, we've got this framework, which we call the Goldie command deck.
So you see it, you steer it and you set it. So this means that you don't need to crawl under the hood anymore. You don't need to manage your terminal. You can run your whole AI from one command deck, which is short enough and fast enough to start using immediately.
So move number one is see it. So you've got one screen that shows you everything. Is it running? What's it doing?
Every past conversation, how many tokens you're using. You don't need to guess anymore. So for example, we open this up, we can see our previous chats here. We can see how many messages we've sent.
We can see all our models. We can see all the logs, et cetera. We can see the chat way easier. And then you can steer it, right?
So the great thing about this is if you need to change any settings, add a key, switch to AI's model, turn skills on and off. You can do that all by clicking. You don't need to use terminal command anymore. So you can switch between all of these.
And also if you want to update the latest version or restart the gateway, you can do that too. You're going to see what version you're on too. And then you can manage your MCPs. You can manage your channels.
You got your plugins, your skills, and you can switch between categories inside the skills as well. You got your tool sets, the browse hub as well. So this is really cool because you can actually search for skills. So if we search for skills here, like social media, for example, you can see loads of different skills that we can actually just one click install here.
So this is great because you basically have a catalog of skills that you can look through, decide which ones you want to use, and then one click install them. And then if you ever want to manage them or delete them, you can go to all your skills here and you can toggle them on and off as well. So it makes it much more accessible to give your AI agent new skills in one single place. The same for your scheduled tasks.
If you've got any scheduled tasks set up, you can manage them here and delete them or organize them. You can also create new ones. So if we click on create here, we can add a new scheduled task. It just makes it way easier to manage in a nice UI.
And then the model section. I think this is really useful because as so many new models, even over the last week, so many new models have come out, like you've got Gemma 4, you've got Claude Opus 4.8, Minimax M3, you've got Grok Step, 3.7 Flash, et cetera. And you can change all of them. And then you can decide which is the main model and which is going to do the subtasks.
Now, why is that important? Because you might, for example, have one main model that's a free one, and then you might have a second backup one that's a free one as well. And then if you're using free APIs, you can organize between them without being rate limited. And then the final thing is set it, right?
So you can schedule jobs and connect your apps from a form and then set it on autopilot and walk away. And it just runs itself, which is great. So if you set up a scheduled task, you can walk away and you know your scheduled tasks are just going to get done. Now, some people think this is pretty technical to do, but you see, for example, like Mike inside AI Profitable, he's built his own mission control.
You see a bunch of other people have built their own mission control and agent operating systems too. So it's getting easier and easier to manage AI agents. And the barrier is no longer technical. If you can text a friend, you can manage an AI agent.
So some people used to think like managing an AI means editing code files and using a terminal. Not anymore. It's a browser-based dashboard with buttons, toggles, and forms. So if you can use the settings screen on your phone, you can do this.
Other people say, I can never tell what my AI is actually doing. So you can manage everything. You can see everything that's been done recently, including the logs and the sessions and all the models. And so you have every conversation and everything inside one place, and it's all searchable too.
And then other people say setting up automations, it needs a developer. Doesn't it? Actually, no, because you can set up your, just inside the Cron section here. Cron is basically, stands for Cron job.
It's a scheduled task. And so you can manage all of them across all your different agent profiles. So if you've got multiple Hermes profiles as well, it just became a lot easier to manage all of those in one place as well. And so what does the dashboard let you run?
You can see its status. You can see all the settings and API keys. You can manage all your conversations, all your scheduled jobs. You can see the tokens you've used and the skills, tools, and the health of your AI agent.
And so like the old way was a wrench and terminal way, right? Editing code, filed settings, typing commands into a black terminal. No easy way to see conversations or tokens. And one typo could break your whole setup.
That was the honest truth. Now, the new way is like you've got a browser, you point and you click. Every setting is a form or toggle. You can see the status, chats, logs, and costs.
You can schedule jobs and connect apps and you can run it on your machine. So it's the same powerful agent, but you've now got a control panel a human can actually use. And so you've got every control in one screen. And we've actually plugged it into the agent operating system now.
So you don't even need a separate app. I've actually wired the whole Hermes dashboard into agent operating system inside the manage app. So you can switch between, for example, Claude, OpenClaude, Hermes, and then you can switch on here. And then also you've got your memory down here, which means that you can connect everything to one vault with all of your memories and all of the information and all of your knowledge previously, so that your AI agents have context instantly too.
So just to recap, Hermes used to be managed with code file settings, terminal commands, it was a wall and a barrier for most people, people intimidated by using Hermes agent. But I think if you look at this update and you look at the desktop application that's just come out as well, they are making it easier and easier to adopt. Which is why I think Hermes is growing so fast. So if you look at the growth of Hermes agent, they've become the most popular AI agent in the world.
Why is that? It's because it's easier than ever before and non-technical people can now use it. And it's smoother when you actually start using it too. OpenClaude was always an option, but OpenClaude was never that smooth to use in the first place.
So the new web dashboard turns all of this into a clean browser command deck with your settings, your keys, your conversations, your scheduled jobs, your skills, your tokens, all point and click all on your own machine. And the command deck framework is just three simple moves. You see it, you steer it, and you set it. And the best part is we've wired it into the agent operating system.
If you want to get that, you can get it inside the AI Profit Boardroom. Link in the comments description or go to the AIprofitboardroom.com. This is my AI automation community for learning how to save time, grow and scale, and also connecting with awesome people who are using AI agents just like you. You can see that it's a very active community with loads of cool people posting awesome stuff.
And there's always people online 24 seven to help you. Inside the classroom, you can get access to all my best trainings. There's also a beginner to expert AI course here, plus new daily updates. So we have the agent operating system.
You can get the zip file down there. We've got a video tutorial, all the prompts of Roadmap for implementing it and new daily tutorials like you see right here, plus you can jump on four weekly coaching calls where you can share your screen, ask questions in real time, and inside the map, you can actually get access and meet people in your local city who are using AI agents and AI automations just like you. So feel free to get that. Link in the comments description or just go to the AIprofitboardroom.com.
Thanks for watching. Cheers. Bye bye.
More episodes