Obsidian for AI-Assisted Notes
Obsidian is a note-taking app built on plain local Markdown files rather than a proprietary database. That one design decision is why it became the default substrate for an AI-assisted second brain: any coding agent, script, or model can read and write your notes directly as files, with no export step and no API to authenticate against. By 2026, Obsidian had grown past "note-taking app" into what practitioners describe as a cognitive infrastructure layer - a substrate AI agents think, remember, and reason on alongside you, backed by 1.5M+ monthly active users and a five-year-old plugin ecosystem.
Why Plain Markdown Files Win
What it gives you
- • Any AI tool can read/write notes as ordinary text files - no plugin required
- • Local-first - notes work fully offline, and nothing leaves your machine unless you choose a cloud model
- • Model-agnostic - use Claude, GPT, Gemini, or a fully local model on the same vault, and switch freely
- • Git-friendly - a vault is just a folder, so version history and diffing work like code
Trade-offs
- • No built-in server-side search across huge vaults without a plugin
- • Sync between devices requires a paid add-on or a third-party tool (e.g. Git, Syncthing)
- • Plain files mean no enforced schema - structure is a discipline you keep, not a constraint the app enforces
The Official CLI
Since February 2026, Obsidian ships an official command-line interface exposing 100+ commands - search, note creation, daily-note management, content appending - all scriptable from the terminal. This is what turned Obsidian from "an app an AI can read files from" into "an app an AI agent can operate directly," the same way a coding agent operates a codebase through a shell.
AI Plugins: RAG Over Your Own Vault
Community plugins add semantic search and chat directly inside Obsidian. The best-known, Smart Connections, embeds your notes and lets you chat with your entire vault - it searches all notes, finds relevant passages, and synthesises an answer with the source notes linked, the same retrieval pattern covered in RAG, applied to your own writing instead of a company's documentation.
Obsidian vs Notion vs Plain Folders
| Tool | Storage | Best for |
|---|---|---|
| Obsidian | Local Markdown files | AI-agent-operated PKM, privacy-sensitive notes, technical users comfortable with plain files |
| Notion | Proprietary cloud database | Team wikis, structured databases, built-in AI features with no setup |
| Plain folders + Git | Local Markdown files, no app | Developers who want zero lock-in and are fine without backlinks or a graph view |
Getting Started
- Start with one folder of plain Markdown files - resist building an elaborate tag or folder taxonomy before you have enough notes to need one
- Install a single AI plugin (Smart Connections is the common starting point) rather than several at once
- Point a coding agent at your vault folder the same way you would a code project, and ask it questions about your own notes
- See Karpathy's LLM Wiki Pattern for a specific, structured way to let an agent maintain the vault for you rather than you maintaining it by hand
Checklist: Do You Understand This?
- Can you explain why plain Markdown files, not a proprietary database, is what makes a vault AI-agent-friendly?
- Do you know what the official CLI added that file access alone didn't already provide?
- Can you state one concrete trade-off of choosing Obsidian over a tool like Notion?