Welcome @Dana Green !

welcome @Robert Alfaro . Greetings from Amsterdam!

Camp 2 here, slowly sliding into 3. I never really decided Obsidian would be just my database. It kind of happened on its own. At some point I noticed I barely opened the app anymore and everything was running from the terminal. My setup is pretty simple. Markdown vault on disk, Claude Code pointed at the root, and a CLAUDE.md on top that tells it how the vault works and which folders to stay out of. The vault is basically just memory the agent reads and writes to. What made me trust this approach is that Anthropic built their own agent's memory on markdown too, so it's not some hack I came up with. The honest downside: it falls apart when the vault gets big. Throw an agent at a couple thousand notes and it burns tokens and loses track of what it was doing. What helped me was adding an index.md in each main folder so it follows a map instead of crawling everything, and using separate subagents for the repetitive stuff like cross-linking and weekly reviews so they don't clog the main thread. The thing I keep going back and forth on is the new Obsidian CLI in 1.12. kepano said something like "now any AI agent can use Obsidian" and I think that actually changes the question. Maybe the agent shouldn't see the vault as a pile of files at all, and should go through Obsidian's own index and links instead. So no, Obsidian isn't really my interface anymore. Feels more like it's turning into the API. Curious what everyone else is running. What's the part of your setup that keeps breaking?

Lets unlock that course together my friends

I appriciate you thanks!;)

Great feedback. Thank you 🙏

Well done mate!