01What AgentOS Is and How It Works
AgentOS is a system that helps you control and manage all your AI agents in one place. Think of it as a dashboard where you can see each agent, give them tasks, and watch them work together. Instead of using separate tools for each job, AgentOS brings everything together.
In this episode, Julian explains how AgentOS can host a full company of AI agents that work 24/7. For example, the system includes Paperclip, which creates a team of agents that build things for you. You can also add tools like Open Design and use them within the same dashboard.
The goal is to go from an idea to a finished product with just one click.
02Main Features of AgentOS
AgentOS has several key features that make it powerful. First, it includes Paperclip, which lets you create a company of agents. These agents work together without stopping.
Second, it has an AI Agent Mastermind where agents talk to each other, come up with ideas, and add them to a pipeline. You can then build those ideas with one click.
- Paperclip: Build a team of agents that work on projects around the clock.
- Agent Mastermind: Agents discuss ideas and add the best ones to your pipeline.
- Customizable Dashboard: Add your own logo, tabs, and tools like SEO workflows.
- Loop Engineering: Two agents work together, one builds, the other judges, and they loop until the job is perfect.
- Local AI Support: Connect local models like Qwen to run privately and for free.
03How to Set Up AgentOS Step by Step
Setting up AgentOS is straightforward. Julian shares that you can get the full system from the AI Profit Boardroom. Here is the process he describes.
- Download the zip file from the AI Profit Boardroom. It includes the AgentOS system, a video tutorial, and the daily update file.
- Install the system on a VPS (like Hostinger) or a local computer. Colby used a VPS with Hostinger and connected Hermes Core and OpenCore.
- Use the update.md file from the zip to keep your system up to date. Apply it daily to get the latest changes.
- Ask Claude, Codex, or another model to customize the system for you. For example, tell Claude, "Add a tab for each client in the SEO section." Claude will build it.
- Test the system by running a simple task, like generating SEO content. Make sure everything works as expected.
04How to Use AgentOS for SEO and Other Tasks
Using AgentOS for a real-world job is simple. Julian shows how to use it for SEO for real estate clients. First, open the SEO section in the AgentOS dashboard.
You will see a place to plug in a keyword and a case study. The system then generates content for your clients.
To handle multiple clients, ask Claude to add a separate tab for each one. For example, you can have one tab for Client A and another for Client B. Each tab uses its own keyword and case study.
You can also set up custom prompts for each client. If all clients are in real estate, you can use similar prompts but make the content unique.
Another use is building a to-do list app. Just ask Hermes, "Create a to-do list app for me." The agent will build it and store it in the dashboard. You can also use the Loop Engineering feature.
For that, tell one agent to build something and another agent to check the work. The judge agent gives feedback until the task is done right.
05Results and Comparisons Shown in the Episode
Julian shares several results and comparisons. One example is Amanda, who built an SEO-optimized blog running locally with Qwen 330B on a Mac Mini M4 Pro. Julian notes this is impressive because older Qwen versions were slow on his own Mac Studio.
This shows AgentOS works well with local models.
Another comparison is between the two versions of Hermes desktop. The official version from Noose Research is smoother and simpler. The open-source version from GitHub (Fata) has more features but is buggier.
Julian says if you want ease, use the official one. If you want power, use AgentOS instead.
Julian also talks about the Loop Engineering system. It uses two models: one builder and one judge. The judge checks if the task is complete.
If not, it gives feedback and the builder tries again. This can improve output quality. OpenRouter has a similar fusion model, but AgentOS gives you more control.
06Tips, Limits, and Mistakes to Avoid
Julian gives several important tips. First, avoid relying on one single model. When the Fable 5 model was pulled, many people lost their systems.
With AgentOS, you can switch to GLM 5.2, local models, or others. This makes your work anti-fragile, meaning you improve even when things go wrong.
Second, be careful when installing skills from GitHub. Julian warns that some skills have bad intent. His three-step process: read the skill file yourself, have Claude check it, and then ask Claude to create a custom version instead.
Never install a skill you don't trust.
Third, focus on one automation at a time. Don't try to build everything at once. Pick one task, like lead generation or emails, and work on it for a week.
Use Hermes to guide you through the steps.
Finally, update AgentOS daily using the update.md file. Chris asked about updates, and Julian replied within 11 minutes. Keeping your system current prevents bugs.
07Who AgentOS Is For and Why It Matters
AgentOS is for anyone who wants to use multiple AI agents together in one system. If you are a beginner, Julian recommends starting with NA10 to learn the basics. But if you already have experience with AI agents, you can skip straight to AgentOS.
It matters because it gives you complete control over your AI work. You are not tied to one model or one tool. If a model disappears, you swap it out.
If you want to add a new feature, you ask Claude to build it. This flexibility saves time and makes your work more reliable.
Julian also emphasizes that AgentOS is lightweight. You can run it on a MacBook Air, MacBook Pro, or Mac mini. Stan asked about budget Macs, and Julian recommended a Mac mini for a permanent setup or a MacBook Air for portability.
The system works on Windows too, but Mac is more common.
08How AgentOS Uses Paperclip and Obsidian
Paperclip is a feature inside AgentOS that lets you create a team of agents. These agents work together on projects. For example, they can build an app or generate content.
You see their results in the AgentOS dashboard.
Keith asked if one Paperclip company can have a special Obsidian vault. Julian explains that you can set this up by giving each agent custom instructions. Tell the agent to log everything in a specific Obsidian folder.
When pulling context for a project, the agent focuses on that folder. This keeps each project separate and organized.
This approach works well for teams that have different goals. Each team gets its own folder, and the agent only updates that one. The rest of the Obsidian system stays clean.
It is a simple way to manage multiple projects inside one AgentOS.
