01What Is Hermes Jarvis and How It Works

Hermes Jarvis is a voice-controlled interface for the Hermes AI agent. Instead of typing, you speak a command out loud, and the agent goes off and does the work for you. It then reads the results back to you in a natural voice.

The key difference from typical voice assistants is that Hermes Jarvis is wired directly to the real Hermes agent, which has tools to control your computer, browse the web, write files, and more. So when you say, "Schedule research on AI news daily at 3 PM," the agent actually creates a cron job and saves the results to files.

In this episode, Julian Goldie demonstrates how you can use Hermes Jarvis to create notes, schedule tasks, and even build simple apps, all by speaking.

02Main Features of Hermes Jarvis

Hermes Jarvis isn't just a voice chat. It's a full hands-free way to trigger the Hermes agent's capabilities. Here are the main features shown:

  • Voice control of the agent using plain English sentences.
  • Read-aloud responses after the agent finishes a task.
  • Multiple voice options, including MiniMax M3, that plug into the Hermes agent.
  • Schedule tasks like daily research and save logs.
  • Screen and browser control (opt-in) for actions like typing notes or opening files.
  • Integration with Hermes tools for searching, file writing, app building, and more.

The whole flow is: you say it, it does it, you hear it.

03Setting Up Hermes Jarvis

Getting Hermes Jarvis running is straightforward. Julian explains two main paths. Here’s how to get started:

  1. Get the Agent OS (through the AI Profitable Auditorium) to have Jarvis pre-configured with the interface shown.
  2. Or build it yourself with Hermes directly.
  3. Set up the right API keys.
  4. Connect a voice model like MiniMax M3 to the Hermes agent.
  5. Make sure you have permission prompts enabled for computer control.

If you go the DIY route, you need to handle the API connections yourself. The Agent OS simplifies this with a ready-made tab.

04Using Hermes Jarvis in Real Examples

In the episode, Julian shows several of his own commands. First, he says, "Hey, can you type in a new note inside my notes and just give me a list of automations I could create?" The agent then creates a note with a list of automation ideas across categories like content, social, video, audio, SEO, business, AI, dev, productivity, and smart home.

Next, he says, "Schedule researching the latest AI automation news daily at 3 PM." The agent creates a cron job that saves results to a daily research log and individual JSON files by date.

He also says, "Build out a to-do list app that's just simple HTML and then open it up for me." The agent goes into the background and codes the app, taking a couple of minutes.

You can also just have a live conversation with the agent in the chat section, not just voice commands.

05Results and Comparisons from the Episode

Julian compares Hermes Jarvis to regular voice assistants like a chat voice mode. He notes that most voice AI only talks back and describes how to do things—it can't actually do them. Hermes Jarvis is different because it acts.

For example, the note creation and scheduling tasks were completed automatically. The to-do list app was built and opened without further input.

He emphasizes that it's not instant; the app building took a couple of minutes. But that time is background time while you do something else.

He also mentions using a new model called Nemetron (Nvidia Nemotron 3 Ultra) as the underlying agent model, which just came out that day.

06Tips, Limits, and Common Mistakes

Julian gives important advice for using Hermes Jarvis safely. Since the agent can do real things, treat it like a keen new assistant that takes you literally. It will act on what you say, so be specific.

  • Give permission only when you're ready. Computer use and screen control are opt-in—the agent will ask for approval first.
  • Don't expect instant results. Multi-step tasks like coding can take minutes, but you can walk away.
  • Use plain English. No special wording needed, just a clear command.
  • Double-check schedules. For example, the daily cron job saves files—make sure you know where they go.
  • Avoid giving dangerous commands. The agent will do what you say, so think before you speak.

Also, voice AI isn't a gimmick here—it's wired to real agents that can research, build, and save files.

07Who Should Use Hermes Jarvis and Why It Matters

Hermes Jarvis is for anyone who wants to delegate actual work to AI, not just ask questions. If you're a founder, marketer, developer, or creator, you can speak a command and get a task done: research competitors, write a brief, build a landing page, or find AI news—while you do something else.

It's also fun to play with and makes multitasking easier. You can keep your hands and eyes busy while the agent works in the background.

Julian says Hermes is by far the most powerful AI agent he's used, and Jarvis makes it even more accessible. The Agent OS includes it, plus studio tools, sessions, MCP support, and more.

If you want the full setup, you can get it in the AI Profitable Auditorium, where Julian personally answers questions and provides daily updates.