AI Second Brain
Obsidian AI: The Setup That Turns Notes Into Answers
Obsidian AI explained: Obsidian has no built-in AI, so here are the plugins and agent setups worth using, how they differ, and which one fits your notes.

Obsidian AI isn't a feature you switch on. Obsidian is a local, plain-text markdown app with no built-in AI, so "Obsidian AI" means one of three setups you add yourself.
Those are plugins that add chat and related-note search inside Obsidian, MCP servers that let AI assistants use your vault, or agents like Claude Code that read and rewrite the notes themselves.
I use all three levels on my own vault, which holds the notes that run my business. Here's what each one is good for, the tools worth installing, and how to pick without drowning in plugins.
Why Obsidian is so good for AI
Obsidian's design happens to be exactly what AI tools need. According to Obsidian's own site, notes are stored privately on your device as plain markdown files in open formats, with no lock-in, and the app collects no telemetry.
That matters for three reasons:
- Files are readable by any tool. An agent doesn't need an export or an API to understand your notes.
- Links are structure. Wikilinks and backlinks give an AI explicit relationships instead of guesses.
- You own the result. When an AI reorganizes a note, you can see the diff, keep it in Git and roll it back.
Obsidian is also free without limits for personal and commercial use. Sync costs $4 a month billed annually and Publish $8 a site, both optional, per the pricing page.
The three levels of Obsidian AI
| Level | What it does | Examples | Best for |
|---|---|---|---|
| 1. In-vault plugins | Chat with notes and surface related notes inside Obsidian | Smart Connections, Copilot, Smart Composer | Finding and asking while you write |
| 2. MCP servers | Expose the vault as tools to AI clients | MCP Connector, mcp-obsidian | Using the vault from Claude Desktop, Claude Code or Cursor |
| 3. Agents on the files | Read, write, restructure and maintain notes | Claude Code, Codex | Running a real AI knowledge base |
Most people start at level 1 and stop there. The big gains come from level 3, where the AI stops being a search box and starts maintaining the vault.
Level 1: the best Obsidian AI plugins
Smart Connections is the plugin I'd install first. It shows notes related to the one you're working on, in a list and a graph view, using an embedding model that runs locally. There's no setup and no API key, so nothing leaves your machine for this feature. It's actively developed and has over 5,000 stars on GitHub.
Copilot for Obsidian adds a chat panel where you can ask questions about your notes using the model provider you configure. It's one of the most popular AI plugins in the community.
A few more in the same family: Smart Composer for AI writing help with note context, Gemini Scribe if you use Google's models, and Agent Client, which lets you chat with Claude Code, Codex or Gemini CLI from inside a vault.
The limit of level 1: these plugins answer questions and suggest links, but they don't maintain the vault's structure for you.
Level 2: connect Obsidian to AI assistants with MCP
MCP servers turn your vault into a set of tools any MCP-compatible assistant can call: search notes, read files, append to a daily note, follow backlinks.
The standout is the MCP Connector community plugin. Its server runs inside Obsidian on your machine only, offers dozens of vault tools, includes on-device semantic search, and gives each connected client its own token, so Claude Desktop can see a small safe set of tools while Claude Code gets more. The plugin page documents the full setup.
I compare the MCP options, and how to connect them safely, in Obsidian MCP.
Level 3: Claude Code as your vault's maintainer
The most capable Obsidian AI setup doesn't need a plugin at all. Open a terminal in your vault folder, run claude, and Claude Code can read every note, create pages, fix links and restructure folders.
This is the setup behind Andrej Karpathy's LLM wiki pattern: raw sources in one folder, an AI-maintained wiki in another, and a rules file that tells the agent how to keep it consistent. My guide to Karpathy's LLM wiki walks through it.
Two small plugins make level 3 more comfortable:
- Claude Code IDE lets Claude Code see which note is open and what text you've selected, through a local, read-only bridge. You connect it with
/idein Claude Code. - Obsidian CLI lets Claude Code run Obsidian commands, like opening notes or renaming files with automatic link updates.
One privacy point the Claude Code IDE plugin's FAQ states plainly: Claude Code sends your prompts and the context it reads to Anthropic's API, with or without the plugin. Keep truly private notes in a separate vault.
Which Obsidian AI setup should you use?
- You mainly write and want better recall: Smart Connections, plus a chat plugin if you like asking questions.
- You live in Claude Desktop or Cursor: add the MCP Connector plugin so they can search and update the vault.
- You want a knowledge base that maintains itself: run Claude Code in the vault with a CLAUDE.md schema and skills for ingesting and linting.
- You manage knowledge for a team or clients: level 3 with shared conventions, Git and a review step.
The mistake to avoid
The most common Obsidian AI failure isn't a bad plugin. It's letting AI write confident summaries that nobody can trace back to a source. Six months later the vault is full of polished claims with no origin.
The fix is structural: keep raw sources immutable, make every summary link to its source, and treat numbers as claims that need evidence. That's the rule I build into every AI Brain. If you want to set this up yourself first, start with how to build an Obsidian second brain with Claude Code.
Frequently asked questions
Does Obsidian have AI built in?
No. Obsidian is a local, plain-text markdown editor with no native AI. AI features come from community plugins such as Smart Connections and Copilot, or from external agents like Claude Code that work on the vault's files.
What is the best AI plugin for Obsidian?
It depends on the job. Smart Connections is best for surfacing related notes with local embeddings, a chat plugin like Copilot suits quick questions inside Obsidian, and an MCP connector is best when you want Claude Desktop or Claude Code to use the vault as a tool.
Is it safe to use AI with my Obsidian notes?
Obsidian stores notes locally and doesn't collect telemetry, but any cloud AI you connect receives the content you send it. Local-embedding plugins keep search on your device; for sensitive notes, use a separate vault for AI work.
Can Claude read my Obsidian vault?
Yes. Claude Code can read and edit a vault directly when you run it inside the vault folder, and Claude Desktop or Claude Code can also connect through an Obsidian MCP server.
Sources
Christian Monge
Builds AI second brains with Obsidian and Claude Code, co-founded the competitive-intelligence SaaS Pengu Insights, and trains teams in English and Spanish from Madrid.
More about meAI Brains
Want this running on your own material?
I build Obsidian + Claude Code brains around your documents, calls and research, with an evidence ledger so every number has a source.
See the AI BrainsKeep reading

AI Second Brain
AI Note Taking App: How to Choose One You Keep
How to pick an AI note taking app: the four jobs they do, why file ownership decides the long game, and the setup that makes notes usable by AI later.

AI Second Brain
Karpathy's LLM Wiki: How to Build One in Obsidian
Karpathy's LLM wiki explained: why a persistent markdown wiki beats RAG, the three layers and three operations, and a step-by-step build with Obsidian.

AI Second Brain
Obsidian MCP: Connect Claude to Your Vault Safely
Obsidian MCP explained: the servers that connect Claude to your vault, how to set each one up, when you don't need MCP at all, and how to keep notes safe.