Claude Desktop + Obsidian + Agent OS AI Memory Systems!

Julian Goldie (0:00)

Today, I'm going to show you how I set up Obsidian and Claude together, so that you never have to be the messenger between your AI agents again. They can work together, they can work on different projects. And this is based on a question that we've had from Andy R inside the AI Profit Boardroom, who says, I'm running into a few AI workflow struggles and would love some direction.

So for projects in Claude, I use Claude Desktop as the architect, this is from Andy, and Claude Code for Development. By end up getting and acting as the messenger between them, context gets lost. Claude Desktop can drift into a new direction and have to keep relaying decisions back and forth. Between Claude Desktop and Claude Code, this is. He says he's had a similar experience with Chat Cheapity and Codex, a worker thread and a planning suggestions thread, with him stuck in the middle managing context, which is no fun for anyone. He's also run into hallucinations and agents going off in their own direction. The biggest issue for him is memory, preserving past decisions, managing context, and keeping a project coherent over time. Now, he's interested in the Agent OS, but I'm going to first of all cover Claude, and then we'll cover the Agent OS setup that I recommend for you. So let's get straight into Claude. So if you're using Claude Desktop, I actually don't think that you need to switch between Claude Desktop and the normal Claude Code. For me personally, in pretty much everything that I do inside Claude, I only use the Claude Code section, right? So you've got Claude Code over here, and then you have Chat and Co-Work over here. Now, I would recommend that if you're switching between them, and maybe for example, it sounds like you're using Claude Code inside the Desktop for development, and then Claude Code Desktop as the architect. For me, I would just keep it simple and just focus on one. I saw that you're having the same problems with Chat Cheapity and Codec. So work a thread and a planning suggestion thread with him, with you stuck in the middle managing context. So for me, if I'm using Chat Cheapity these days, I just use the app. I think it's far simpler just to focus on the app and stick to that. And then for example, I can have a project set up, like for example, the Agent OS. And if I need a new chat, I can click on the new chat, but it's all working on the same project. So that way, we keep everything in one folder for each project. So we've got the Agent OS, for example, we have the AR Profit Boardroom website in another folder, and everything is separated into separate projects. We can have separate chats between them, and I don't need to be a messenger between any of them. This also fixes all the issues that you're having with hallucinations. So if your agents are like hallucinating inside Chat Cheapity or hallucinating between Chat Cheapity and Codex, I would just stick to Chat Cheapity app to avoid that. It will fix the problem for you. So now let's talk about the biggest issue, which you were saying was memory, preserving past decisions, keeping a project coherent over time and using the Agent OS system. So the Agent OS system, if you're going to use that for memory instead, you would use the memory section. Now for me personally, the way that I've set it up is in the recent section, I have like decisions logs, I have daily updates, pretty much everything that's important, that's done by my AI agents is plugged into a new memory inside our system. Now, if you want to do this, here's the system that we use, right? And I actually just build this custom for you, just to help you as much as we can.

So the first thing that you want to do is choose projects inside the memory tab. Get the latest version of the Agent OS, and then you can go to the memory tab as you can see right here. And then you can click on memory and projects, and you can import a project. So you want to choose your existing project, so you can click import a project, paste a path of your project's folder, and click preview import.

And then it can do it like that, as you can see right here. So we can add the project details locally for this. Another option is that you can just say, okay, import the project that I currently have into my Agent OS, and you can ask any agent to do that for you. And then from here, you would explain what you want to build.