01What Is AgentOS and How Does It Work?

AgentOS is an agent operating system that lets you run multiple AI models together. It has a memory system that works across all your AI agents. You can switch between models like Claude, Hermes, and Quen 3.8 Max with one click.

When a new useful model drops, you can plug it into AgentOS right away. The system remembers what you learned and uses it for future tasks. It also has a special command, forward slash learn, that turns a guide into a skill the AI can use on its own.

This makes AgentOS flexible and easy to update without coding.

02Main Features of AgentOS You Should Know

AgentOS has several key features that make it different from using a single AI. Here are the main ones Julian showed:

  • Memory system: The AI remembers information across sessions and agents, so you don't have to repeat yourself.
  • Multi-model support: You can connect Claude, Hermes, Kimi, GPT 5.6, and Quen 3.8 Max. Switch models in one click.
  • Forward slash learn command: Give the AI a link to a guide, and it creates a skill file (a skills.md file) that it can load later.
  • Forward slash skill command: Load a saved skill by typing its name, like forward slash skill agent hyphen operating hyphen system.
  • Integration with Unreal Engine: AgentOS works with Hermes to automate 3D graphics for games and apps.

03How to Set Up AgentOS Step by Step

Setting up AgentOS is straightforward. Julian explained the process using Kimi or Claude. Follow these steps:

  1. Get the AgentOS zip file from the classroom inside the AI Profit Boardroom community.
  2. Open Kimi Work (a desktop app) or Claude Code. If you use Kimi, open the app and give it the zip file.
  3. Ask the AI to install the AgentOS for you. For example, say: 'Set up the AgentOS from this zip file.'
  4. Wait for the AI to check that everything works. It will confirm the installation is complete.
  5. Test the setup by trying a simple command, like forward slash learn on a test guide.
  6. If you want to use Fable 5 to modify AgentOS later, just explain the change in plain words and let Fable 5 build it into the system.

04How to Use AgentOS: Examples from the Episode

Julian showed several ways to use AgentOS. First, to teach the AI a new skill, use forward slash learn. For example, go to a guide about AgentOS, copy its link, and type forward slash learn plus the link.

The AI will read the guide, create a skill file, and save it. Julian did this with a guide on DeepSeek and got a skill called open code hyphen Hermes hyphen Kanban.

To load a saved skill, type forward slash skill followed by the skill name. The AI will recall what it learned and apply that knowledge. You can also use AgentOS to build apps.

Julian said he built the whole AgentOS by describing what he wanted and letting AI code it. For a simple company app, he recommended going into Claude Code, describing the app, testing it, and iterating.

05Results and Comparisons Julian Showed

Julian compared AgentOS to using single models like Claude alone. He said the whole point of AgentOS is to use multiple models instead of limiting to one. For manufacturing, he recommended Claude with Fable 5 as the most reliable and fastest choice if you must pick one model.

But AgentOS lets you swap models easily when a better one comes out.

He also compared forward slash learn to similar features from Claude. Julian said Hermes nailed the skill-learning feature the best and that he has not seen anyone else do it as well. The results he showed were a skill file created in seconds after giving the AI a guide link.

The AI automatically linked the skill to relevant memory and showed where the file was saved.

06Tips, Limits and Mistakes to Avoid with AgentOS

Julian shared important tips from his own experience. When modifying AgentOS, expect to spend three to four hours testing and iterating even after the AI builds the change. The biggest mistake is not testing enough.

Julian said testing is the main time-consuming part. He also warned about email outreach deliverability: use a separate domain from your main business domain. If your outreach email score drops, it will not hurt your main business.

For warming up an email account, send five emails a day the first week, then ten a day. Keep open rates above 20%. If they drop below 20%, that means emails are not being delivered.

Use DKIM and DMARC settings, a Google Workspace account with a domain (not a personal Gmail), and keep your outreach text conversational with short subject lines.

07Who AgentOS Is For and Why It Matters

AgentOS is for anyone who wants to use multiple AI models together without switching between different apps. It works for non-technical people because you can describe what you want and the AI builds it. Julian said if he and other non-technical people can use AgentOS, you can too.

It is useful for building apps, creating skills from guides, and automating tasks like 3D graphics with Unreal Engine.

Why it matters: AgentOS makes AI agents more capable because they can remember information across sessions and learn new skills on the fly. You stay updated with the latest models without manual work. For businesses, this means faster development and fewer limits from one AI model.

Julian uses AgentOS daily to improve his own AI tools and said it is the most flexible way to work with AI agents.