01What Agent OS Is and How It Works

Agent OS is a single dashboard that controls all of your AI agents. They all work together, talk to each other, and do jobs for you around the clock. This episode shows how powerful it really is.

The system lets you plug in all your agents so they work as a team. You can customize it, change the UI, and add your own workflows. It comes with mission control, memory, and custom workflows built in.

One community member named Michael connected his Agent OS to his online store. It now creates all of his social media content automatically. Another member built his entire setup with mission control, memory, and custom workflows.

You can see these examples in the episode.

02Main Features of Agent OS

Agent OS gives you a whole system for controlling all your agents. You can run multiple different models side by side on the same task. For example, you can have one tab with Claude and one tab with Hermes working on the same task.

  • Mission control lets you oversee all your agents from one place.
  • Memory is built in so your agents remember past tasks.
  • Custom workflows let you personalize how your agents work.
  • You can sync external tools like eBay listings into Agent OS.
  • The UI is customizable so you can make it your own.

The system is given away inside the AI Profit Boardroom. You can also build your own if you prefer. The main point is you can customize it and make it fit your needs.

03How to Set Up Agent OS Step by Step

Setting up Agent OS is straightforward. Here are the steps shown in the episode.

  1. Get access to Agent OS inside the AI Profit Boardroom or build your own version.
  2. Name your agent or project in the terminal. For example, run "Hermes" to start.
  3. Open a new tab and type "Hermes model" to see available models.
  4. Scroll down and select the model you want, such as Kimi/Moonshot.
  5. Grab an API key from your subscription plan. For Kimi, go to the console and click "Create API key".
  6. Copy the key, paste it into the terminal, and select the model (like Kimi K3). The model is now ready to use.

Once set up, you can add multiple agents and compare them side by side. The episode shows a six-step process that gets you running quickly.

04How to Use Agent OS With Real Examples

The episode shows several real examples of using Agent OS. One community member named Michael set up an Agent OS promo machine. It syncs his eBay listings to Agent OS, and then creates social media content based on those listings.

He just signed up, set it up, and he was good to go.

Another member named Lois built their own Agent OS with mission control, Obsidian, Claude Code, and Paperclip. They shared their setup inside the community. This shows how you can customize the system to fit your needs.

You can also use Agent OS to run multiple models side by side. For example, you can test Claude and Hermes on the same task. Open two tabs, give both the same prompt, and compare the outputs.

This helps you find which model works best for your tasks.

The episode also recommends using Goldie Bench to compare models. Inside the compare section, you can see how each model performs. You can test the same prompts on each model and pick the best one.

05Results, Tests and Comparisons From the Episode

The episode shows several tests and comparisons. One big finding is about Claude code. Anthropic removed 90% of the Claude code system prompt for Claude 5 models.

They found no measurable loss on coding evaluations. This means shorter prompts now beat longer ones.

Julian Goldie shares his personal preference. He prefers Fable 5 over Opus 5. He finds Opus 5 goes off track and is very slow.

For coding tasks, Fable 5 is the better choice.

A Gartner survey shows 22% of organizations are reducing entry-level hiring due to AI automation. The focus is moving from entry-level work to agent management. This shows how quickly the industry is changing.

The episode also shows that backlinks and brand searches are key differentiators for ranking websites. Even with expert content from AI, backlinks still take time to build and most people haven't automated them yet.

06Tips, Limits and Mistakes to Avoid

The episode shares important tips and warnings. Here are key points to keep in mind.

  • Simplify your prompts. Shorter prompts now beat longer ones. Remove fluff and let your agents use their own judgment.
  • Add security rules. A community member created ready-to-paste rules like "ask before making changes" and "use approval levels." This keeps your agents working safely.
  • Don't give AI agents email access unless you are comfortable. Set up a test account first and read through the skill file.
  • Back up your website before switching themes. Many people mess up by switching too quickly without preparing.
  • Audit your time before building automations. Focus on one automation per week. You will build 50 automations per year.

Local models are not great according to the episode. If you use them, link them to your Obsidian database for memory. Fine-tuning is a separate topic not covered in depth here.

07Who Agent OS Is For and Why It Matters

Agent OS is for anyone who wants to control multiple AI agents from one place. It is especially useful for business owners, agency owners, and content creators who automate repetitive tasks. The episode shows how community members use it for social media, eBay store management, and SEO.

The system matters because the industry is changing fast. Entry-level jobs like data input, spreadsheet management, and content creation are being automated. The focus is moving to agent management.

Learning Agent OS now positions you as an expert in this growing field.

Julian Goldie recommends documenting what you build daily. Share on YouTube, Twitter, or social media. This positions you as an expert and leads to client work.

Content marketing is the best way to get clients in this new market.