AI‑first PKM setups

The newest personal knowledge management fad is not really about notes. It is about turning a folder of Markdown files into a private operating system for thought. That is why Obsidian keeps showing up at the center of the trend. The app stores notes as plain-text Markdown files inside a local “vault,” which is just a folder on your computer, and it lets users reshape the app with thousands of plugins and an open API (obsidian.md 1) (obsidian.md 2). That design matters more in the age of AI than it did in the age of backlinks and graph views. If your notes are ordinary files, an AI model can read them, sort them, summarize them, and write new ones without waiting for a note app company to expose the right feature. Obsidian explicitly supports importing Markdown files by simple drag and drop or by moving files directly into the vault folder, which is why users now describe these vaults as “command centers” that absorb keep.md files, clipped documents, and machine-generated drafts (obsidian.md 1) (obsidian.md 2). That is the real shift behind the recent burst of social posts. People are no longer treating PKM as a place to manually file ideas. They are building small local systems that feed notes to models, then ask those models to clean inboxes, extract tasks, draft summaries, and connect related material across the vault. Obsidian’s community plugin system was built for exactly this kind of extension. The company’s own documentation says community plugins can add support for third-party services and new file formats, while warning that they run third-party code on the user’s behalf. That warning is not a side note. It is the price of all this flexibility (obsidian.md) (docs.obsidian.md). The AI layer itself is splitting in two directions. One path stays local. Ollama, one of the most common tools in these setups, is built to run large language models on a user’s own machine and now offers a broad library of local models alongside optional cloud access (ollama.com 1) (ollama.com 2). The other path uses a stronger remote model, often Claude, but tries to keep the user’s notes under local control. That is where the Model Context Protocol, or MCP, enters the picture. MCP is an open standard for connecting AI applications to external data sources and tools, including local files, and Anthropic pushed it into wider use through Claude Desktop support and one-click desktop extensions for local MCP servers (modelcontextprotocol.io) (anthropic.com). Once MCP existed, Obsidian was an obvious target. Open-source projects quickly appeared to let Claude read, search, write, and manage notes inside a vault through natural-language commands. That is not an official Obsidian feature. It is a community-built bridge. But it fits the moment perfectly because it preserves the thing users care about most: the notes remain files they can inspect and move, not entries trapped inside someone else’s database (github.com) (obsidian.md). This is also why the trend has a sharper edge than the usual productivity boomlet. The excitement is not just about convenience. It is about control. Cloud-first note apps still win on collaboration and polish, but they are harder to bend into AI workflows that mix local documents, custom scripts, and self-hosted models. Obsidian wins this round because its core abstraction is boring in the best way. A vault is a folder. A note is a file. In 2026, that looks less like nostalgia than infrastructure (obsidian.md) (obsidian.md).