NEW Hermes AI Voice Agent Changes Everything

Added: 2026-06-06

[00:00:00]New Hermes AI voice agent changes everything. What if you could just talk to your AI agent and it actually did things, not just set a timer, not just answer a trivia question, actually control your tools, your agents, your entire workflow with your voice? And what if you didn't need to be a developer to set it up? Nobody's really talking about this yet, but today I'm going to show you exactly how it works and why this is one of the most powerful setups I've seen this year. Hey, I'm the digital avatar of Julian Goldie. I'm here to help you actually learn and use AI tools in your work so you can build faster and get more done with less effort. Today we're going deep on Hermes agent, a real-time voice agent that lives inside a full agent operating system. I'll walk you through what it is, why it matters, show you real use cases, then give you the exact steps to get it running yourself. Let's get into it. So first, what even is Hermes agent?

[00:00:50]Hermes is an open-source AI agent built by New Research. It's not just a chatbot, it's a persistent agent, meaning it remembers what it learns, builds its own skills over time, and runs on your own machine. It works across CLI, messaging platforms like Telegram, Discord, Slack, and WhatsApp, and it also has full voice mode, so you can literally just talk to it hands-free. The key thing is that Hermes isn't tethered to one tool or one API.

[00:01:15]It has persistent memory across sessions. It can schedule tasks, it can browse the web, it can run code, it can delegate tasks to sub-agents, and it does all of this while you're just talking to it. The longer it runs, the more it learns about your setup and your workflow. Now, what powers Hermes? The brain I like to plug in is Minimax M3, and this matters because the voice is only as good as the model behind it.

[00:01:38]Minimax is a world-leading AI company with over 200 million global users.

[00:01:44]Their latest model M3 is a genuinely frontier model. It has a 1 million token context window, which means it can reason across enormous amounts of information in one session. It's built for coding and agentic tasks. It doesn't just chat, it thinks and acts, and it's natively multimodal. It can generate images, understand video, and handle text all from the same model. MiniMax M3 is also going to be released open source on Hugging Face. So, in the future, you'll be able to run it locally yourself. This is why Hermes doesn't feel like a basic assistant. It feels like you're talking to something that actually reasons. That's the M3 brain behind it. Now, let me show you what this actually looks like in practice. I opened Hermes inside Agent Operating System and clicked into the voice interface. You can switch between different voices and accents right from the UI. So, I tapped and asked it, "Hey, are you Hermes?"

[00:02:35]>> I'm Hermes, MiniMax's AI assistant. What can I do for you?

[00:02:38]>> Tell me a joke.

[00:02:38]>> Why did the scarecrow win an award?

[00:02:40]Because he was outstanding in his field.

[00:02:42]>> What's the most interesting thing you can automate for me?

[00:02:45]>> summarize your daily notes into a single today page with a single command.

[00:02:49]>> Teach me Japanese. And just teach me a few basic conversational words in Japanese.

[00:02:53]>> Sure. Konnichiwa. Hello. Arigato. Thank you. Sumimasen. Excuse me. Hi. Yes.

[00:02:57]Yeah. No. Onegaishimasu. Please.

[00:03:00]>> Now, the bigger picture on why this matters. Think about the old way. You'd have a voice assistant that could set timers. And if you asked anything outside that lane, it would just say, "Sorry, I can't help with that." You'd be back to typing prompts, copying and pasting, juggling 10 browser tabs.

[00:03:14]That's where Agent OS changes everything. Agent OS is an open-source system created by Brian Casel at Builder Methods. It's a system for injecting your code base standards and writing better specs for what's called spec-driven development. It works alongside tools like Claude Code, Cursor, and others. Any language, any framework. The core idea is a structured workspace where your standards, your specs, and your agents all live together. Agent OS helps you discover the patterns in your code base, deploy those standards intelligently based on what you're building, and shape better plans before you write any code. What this means for Hermes is powerful. When you run Hermes inside Agent OS, you've got all your agents in one place. You can see everything you've previously built. You can switch between agents like jumping into Open Claw or any other agent you've set up right from the same studio interface. One system, every tool plugged in. And here's the important part, you don't need to be a coder. The old way required a team of developers to build a setup like this. Now you just tap and get straight into it. One feature that just dropped is skill bundles. You can load multiple skills into Hermes at once. For example, you could install the browser catalog from browser.sh along with a browser-based CLI and load a full bundle of browser use agent skills all at once. You see them right there in the chat, grouped together, switch them all on with a single prompt, and you're ready to go. That saves serious time versus setting up each skill one by one.

[00:04:36]Now, here's where I want to talk about the AI Profit Boardroom. If you want my exact setup, this whole agent operating system with Hermes Mini Max M3 and the voice workflows running on top of it, join the AI Profit Boardroom. The full agent OS setup lives inside there. You get the complete walk-through on how it's built, how the agents connect, and how to wire up your skill bundles the right way from day one. You can use it to run voice automations, control your agents hands-free, and switch between agents from one studio. You can build browser use workflows that navigate sites and fill forms for you. You can schedule tasks to run on their own. You can have it auto-summarize your daily notes into a single today page. You can generate images and video straight from the same brain that's talking to you.

[00:05:21]And we go live on coaching calls where you can ask questions about your own setup as you build it. Everything you're seeing in this video, we go deeper on inside the Boardroom. All right, let's talk installation. For Hermes agent, the install is clean. You run a single curl command from the Hermes agent website at hermes-agent.nowresearch.com.

[00:05:40]For Mac or Linux, it's one line, curl >> https://hermes-agent.nousresearch.com/install.sh pipe bash. Then run Hermes setup to configure it. That's where you connect your model provider and plug in MiniMax M3. Full guide at hermes-agent.nousresearch.com/docs/getting-started/installation.

[00:06:10]For Agent-OS, the repo is at github.com/buildermethods /agent-os and full installation docs are at buildermethods.com/agent-os/installation.

[00:06:24]It works alongside Claude code, so if you've already got a setup going, Agent-OS drops right into it. A few pro tips for beginners. First, don't skip the Hermes setup step. This is where you connect your model, configure your messaging platforms, and set up your tool gateway. Rush past it and the agent won't have the context it needs. Second, use skill bundles. Don't add things one at a time. Group related skills together and load them in a single go. It's faster and keeps your setup clean.

[00:06:52]Third, let it run. Hermes is built to get better the longer it's active. The persistent memory and auto-generated skills mean it learns how you work. Give it a few sessions. Fourth, for voice mode, experiment with different voices and accents from the start. There are multiple TTS providers built in, including MiniMax speech, Eleven Labs, and OpenAI TTS. And fifth, check the Hermes GitHub at github.com/nousresearch/hermes-agent.

[00:07:21]It's actively maintained with regular updates. So to wrap up, what you've got is a real voice agent, not a toy, not a chatbot wrapper, a full autonomous agent that runs locally, remembers everything, connects to your tools, and can be controlled entirely by voice. It's powered by a frontier model with a 1 million token context window that's going open source. It lives inside Agent OS where your skills, your schedules, and every agent all sit in one workspace. And you don't need to be technical to use it. This is the direction everything is heading. Agents you talk to like a person that actually think and do things. If you want the full process, SOPs, and 100 plus AI use cases like this one, join the AI Success Lab. Links in the comments and description. You'll get all the video notes from there plus access to our community of 75,000 members who are crushing it with AI. And if you want my full setup, the complete agent operating system with Hermes and MiniMax wired in, join the AI Profit Boardroom. The whole Agent OS build is in there ready for you. When you get into Agent OS, you're going to run into questions. How do I structure my workspace? How do I wire my agents together? How do I build skill bundles that fit my own use case? That's exactly what the coaching calls are for.

[00:08:35]You get live sessions, full tutorials, road maps, and prompts you can use straight away. You don't have to figure this out alone. Head to AI profitboardroom.com and I'll see you inside.