01What is an agent operating system and how does it work?
An agent operating system is a single dashboard where multiple AI agents live, coordinate, and work together for you. Instead of using AI like a vending machine—one question, one answer—you get a system that connects your notes, goals, tasks, and memory in one place. It runs in your browser and feels like a command center for your work.
The core has three parts: a front end (the dashboard and chat sidebars), a set of agents (AI models with specific instructions), and a memory/data layer (your notes and goals). Everything else is built on top of these three things. In this episode, Julian Goldie shows how he built his own for free using Claude, just by describing what he wanted.
02Main features of an agent operating system
Julian's agent operating system includes a dashboard with a sidebar of different agents you can talk to. Each agent has a specific job, like handling content, business tasks, thinking through decisions, or connecting to your Obsidian vault. It also has a journal, a memory system that remembers everything across conversations, and voice input so you can talk instead of typing.
Here are the key features you can build into your own:
- A dashboard that runs in your browser, with a clean, modern look.
- Multiple agents, like a daily planning agent, content strategist, or business coach.
- A memory system that reads your notes and goals, so agents know you before you start talking.
- Integration with tools like Obsidian, Notion, or Google Drive for storing your context.
- Voice input for hands-free interaction.
- The ability to add new agents anytime by just describing what you need.
03How to build your own agent operating system step by step
Building your agent operating system is easier than you think. You don't need to code. Julian says the skill is knowing what to ask for, not knowing how to code.
Here are the steps he recommends:
- Pick one specific problem in your business or workflow that AI could help with, like writing client emails faster or researching topics quicker.
- Ask Claude to build a dashboard for that task. For example, say: "Build me a personal AI dashboard with a sidebar of agents, each with a specific job. I want it to run in my browser and look clean and modern."
- Start with one agent only, like a daily planning agent, and get it working.
- Once it works, add a second agent for another task.
- Set up a memory layer, like an Obsidian vault or a plain text file, and tell your agents to read it at the start of every session.
- Keep tweaking and improving daily based on your needs.
04How to use it: walking through real examples from the episode
In the episode, Julian shows concrete examples of how he uses his agent operating system. For his SEO content agent, he doesn't start from zero. The agent already knows his audience, tone, and past successes.
He just tells it the topic, and it helps shape the angle, hook, and structure.
His goals agent helps him prioritize work at 7 a.m. without thinking too hard. His video agent can create videos from a single topic description. And with Hermes connected to his Obsidian vault, he can ask questions and get answers from thousands of notes he wrote months ago.
It's like having a second brain you can actually talk to.
Here are other uses he mentions:
- Automating SEO workflows inside one hub.
- Connecting agents to your goals, projects, and priorities.
- Using OMI to take notes automatically and export them to Obsidian for daily memories.
- Asking agents to pull relevant answers from your past research.
05Results and comparisons from the episode
Julian shares what changed after building his agent operating system. He says it felt like sitting down at a control center. His AI wasn't just answering questions—it knew who he was, his goals, and what he'd been working on.
He compares the difference to a filing cabinet versus a personal assistant. One is organized and intelligent, the other is just a pile of folders.
He also says the gap between people who use AI and people who build with AI is widening every month. Systems compound: every agent you add makes the whole thing more powerful. He notes that building your own system is better than buying off-the-shelf tools because you can change and extend it.
Tools you buy are built for everyone, but a system you build is built for you.
06Tips, limits, and mistakes to avoid
The biggest mistake Julian made was trying to build everything at once—full dashboard, voice input, Obsidian integration, memory, goals, and journal all in one go. He says that was wrong. Instead, start with one agent, get it working, then add more.
Another tip is to write good system prompts. A good system prompt makes an agent feel like it was built specifically for you, not generic. Julian also mentions that the memory layer is key.
By default, AI has no memory, so you have to re-explain yourself every time. Fix that by maintaining files that describe who you are, what your business does, your goals, and your key decisions.
Don't worry if you're not technical. Julian says he doesn't code and doesn't understand code. He handled the business side—clarity and feedback—and Claude handled the technical side.
The barrier isn't coding ability; it's the belief that you can do it.
07Who this is for and why it matters
This agent operating system is for business owners, freelancers, content creators, and agency operators—people with no coding background. Julian says people are building these systems right now to run parts of their business while they sleep. If they can do it, you can do it.
Why it matters: the people who are going to win with AI in 2026 are the ones building systems now. Building your own gives you control and understanding. You can add agents when you need them and update memory files when your situation changes.
The system you build this year becomes an unfair advantage you'll still be using three years from now.
08How to get started with a context layer today
The simplest version of an agent operating system starts with a plain text file. Write down who you are, what your business does, who your customers are, what your current goals are, and what you're working on now. Save it somewhere your agent can access it.
Hand it to an agent at the start of every conversation.
That one file is your context layer. It immediately makes the agent more useful than any generic AI chat interface. From there, you can build it up—add more files, connect to Obsidian or Notion, and expand to multiple agents.
Julian says the format matters less than the fact that your context is somewhere the agent can access it.
