I built my second brain with an obsidian MCP
I know “second brain” gets thrown around a lot, but I mean it pretty literally.
I’ve been building a system where Obsidian is not just where I dump notes. It’s where my AI tools learn how to route, remember, and reuse the important parts of my life and work.
The core idea is simple: I don’t want AI to just answer questions from the current chat. I want it to understand the difference between a thought, a person, a project, a communication, a task, an issue, a piece of knowledge, and something that should become long-term memory.
So I built an Obsidian setup around vaults, schemas, routing rules, and MCP access. I’m calling it an Obsidian MCP because that is actually how I’ve been wiring parts of it: agents can inspect the vault structure, decide where something belongs, create notes, validate them, and connect them back to the rest of the system.
Example: I can give it a screenshot of a Facebook comment I wrote, and instead of just summarizing it, it can figure out that it belongs in People, use it to build a persona note for me, capture my tone, my sentence structure, my AI worldview, and what I actually bring into companies.
That is the part I care about.
Not “AI wrote a note.”
More like: AI understood what kind of memory this should become.
The system is still rough, but the direction feels right. Obsidian becomes the working memory layer. MCP gives agents a real operating surface. The notes stop being random markdown files and start becoming structured memory that can compound.
I’m trying to get to the point where my AI does not just know facts about me. It knows how I think, how I talk, what I care about, what projects matter, what decisions were made, and what should not be forgotten.
That feels a lot closer to a second brain than a folder full of notes.