01What an Agent Operating System Is and How It Works

An agent operating system is a system where multiple AI agents work together. They share memories, tasks, and context so they can operate as a team. This allows each agent to know what others are doing and access the same information.

In this episode, Julian Goldie explains that the agent OS runs in a loop. It iterates, tests its own work, and keeps improving. The system is built using Claude Code or other AI agents, and it can be hosted on a VPS for clients.

02Main Features of the Agent Operating System

  • Shared memory: Agents upload memories into Obsidian using the Infinite Context Engine. All agents can pull in those memories and use them.
  • Goal mode: This mode lets the agent OS run in a loop, constantly iterating and improving based on goals you set.
  • VPS access: The system can be set up on a VPS using Cloudflare or Tailscale, so clients can access it from mobile or desktop.
  • Cross-agent context: One agent, like Buzz, can have instant context of every project, website, and even login details.
  • Skill integration: You can install skills from GitHub to replicate websites or perform specific tasks.

03How to Set Up Your Agent Operating System

  1. Get the agent OS template zip file from the AI Profitable community or a similar resource.
  2. Open Claude Code or Codex and create a new chat. Use the command forward slash goal to activate goal mode.
  3. Upload the zip file to the agent and provide the client's specifications. For example, if the client needs a fitness app and a social media app, list those requirements.
  4. Let the agent run in goal mode. It will iterate, test, and improve the system on its own.
  5. Set up the VPS using Cloudflare or Tailscale. Follow examples from the community, like John Meza's setup.
  6. Test the system with the client to make sure everything works properly.

04How to Use the Agent Operating System in Practice

Once your agent OS is set up, you can use it for many tasks. For example, Julian shows how to build websites by installing a GitHub skill into Claude Code. You give it the websites you want to replicate, and Claude uses the skill to build similar sites.

You can also use the system for email personalization. Combine Hunter API with Claude to enrich leads. Claude can write a personalized first sentence for each prospect in a spreadsheet.

If you have thousands of leads, let Claude work on it for hours.

The agent OS also works with Obsidian. Set up folders using the PARA method (Projects, Areas, Resources, Archives). This sorts information by actionability, making it easy for agents to find everything about your brand, projects, and goals.

05Results, Tests, and Comparisons from the Episode

Julian tested Chad Opus 5 and found it very slow. A task that takes Fable 5 about 20 minutes can take Opus 5 one hour. He switched back to Fable 5 to save time.

He recommends using Fable 5 or GPT 5.6 Sol if you run out of tokens.

Codex with voice mode was also tested. It can orchestrate your whole screen using voice commands. But Julian notes it still lags, breaks, and struggles on some tasks.

He believes it will be very powerful once it becomes reliable.

Community member John Meza successfully set up an agent OS on a VPS using Cloudflare. That system can be accessed from a mobile app and desktop, giving clients remote access.

06Tips, Limits, and Mistakes to Avoid

  • Use goal mode for complex builds. A single prompt won't build a full agent OS. It needs to run in a loop to iterate and test.
  • Avoid slow models for big projects. Chad Opus 5 is not fast enough to build anything substantial. Stick with Fable 5 or GPT 5.6 Sol.
  • Organize memories with PARA. Without a clear folder structure, agents will struggle to find the right context. Obsidian works well for this.
  • Test before client delivery. Make sure the system works and the client can access it on the VPS.
  • Be patient with new features. Codex voice mode and other tools are still in development. They will break and lag, so plan for that.

07Who Should Use an Agent Operating System and Why

The agent OS is for anyone building AI automation for clients. If you want to sell AI agents as a service, this system lets you create, manage, and host agents for multiple clients. It saves time because agents share context and memories.

It is also for freelancers or businesses that want to automate repetitive tasks like email personalization, website building, or content posting. The system can be set up on a VPS so clients can access it remotely. Julian recommends starting with one automation per month to see a massive difference in time saved and profits made.