Obsidian AI: how to ask your vault questions, privately

Obsidian doesn't have a built-in AI assistant. To ask your vault questions with AI, you add a community plugin, like Copilot or Smart Connections, or use a separate app that reads your vault. To keep your notes private, pair it with a model that runs on your own computer, through Ollama or LM Studio, so nothing is sent to an AI company.

This guide covers what Obsidian itself offers, the main AI plugins and what changed in them this year, how to set one up fully offline, and the privacy risks to know about. Obsidian's and the plugins' details were checked on 9 October 2026.

Does Obsidian have AI built in?

Not as a chat assistant. Obsidian's changelog up to version 1.14.4 (5 October 2026) adds no AI features to the app itself, and its manifesto puts privacy first: "Data stays on the user's device." Obsidian is free for personal and work use, and makes its money from optional Sync (from $4 a month, billed yearly) and Publish.

Two official pieces do touch AI:

• Web Clipper's Interpreter can summarize or pull details from a web page as you clip it, using a model you choose, including Anthropic, OpenAI, Gemini or a local model through Ollama.
• The Obsidian CLI lets command-line tools, including AI coding agents, work with a vault.

Asking questions across your notes needs a plugin or another app.

The main Obsidian AI plugins in 2026

Download counts are from Obsidian's own plugin statistics on 9 October 2026.

What changed this year (many articles still describe the old versions):

• Copilot 4 is built around an agent that can work on your vault, and it's desktop only. Searching notes by meaning moved to Miyo, a separate free app that runs on your computer. Without it, Copilot's free chat doesn't search your vault on its own; you point it at notes.
• Smart Connections no longer includes chat. It focuses on finding related notes; chat moved to a separate Smart Chat plugin, and connecting other models is part of its Pro plan ($30 a month or $299 a year).
• Smart Composer, once a popular choice for chatting with a vault, says in its own README that it's no longer under active development.

Another fast-growing option is running an AI coding agent, like Claude Code or Codex, inside Obsidian through plugins such as Claudian. These are powerful, but they usually send your notes to a cloud model.

How to set it up fully offline

Here's the documented way with Copilot, the most-downloaded plugin. You need a Mac, Windows or Linux computer with enough memory for a model, ideally 16 GB or more.

1. Install a local model runner. LM Studio or Ollama, both free. Download a model, for example with ollama pull qwen3:8b, and make sure the server is running.
2. Install Copilot from Settings → Community plugins in Obsidian, and turn it on.
3. Add your local provider. In Copilot's settings, add a provider and choose Ollama (address http://localhost:11434/v1) or LM Studio (http://localhost:1234/v1). No API key is needed. Pick your model and press Test.
4. If chat fails, turn on Copilot's "Enable CORS" setting.
5. For search by meaning across the vault, install Miyo and add your vault to it.

Prefer something simpler? Smart Connections works offline with no setup: install it, turn it on, and it starts indexing your notes with its own small model. It finds related notes rather than answering questions.

Privacy and security: what to watch

• Cloud models see your notes. If a plugin uses OpenAI, Anthropic, Google or another online service, the notes it picks for each question are sent there. Our guide on pasting client files into ChatGPT covers what those services do with it. Copilot also warns that some free hosted models it offers "may log or train on prompts."
• Plugins can read everything. Obsidian's own help page says community plugins can read your files and use the internet, and that Obsidian "cannot reliably restrict plugins." For a sensitive vault, it recommends checking a plugin yourself before installing it.
• API costs add up. With your own key, you pay the AI company per use. A local model costs nothing to run.
• Phones are limited. Copilot's agent and Miyo are desktop only. Smart Connections and Smart Second Brain work on mobile.

Which option should you choose?

• You want related notes to show up as you write: Smart Connections, free and local.
• You want to chat with your vault and don't mind some setup: Copilot with Ollama or LM Studio, plus Miyo for search.
• You want a free, open-source chat that stays local: Smart Second Brain or Local GPT.
• You want answers you can check, from your vault and your other files, on a Mac: a separate app like Citenook, below.

Asking your vault from outside Obsidian

Plugins run inside Obsidian and see only your vault. If your work lives in PDFs and documents too, an app that reads all of it can be simpler. Citenook is a Mac app that finds your Obsidian vaults and reads the notes like a folder, along with any PDFs, Word files and other folders you add. It keeps up as notes change, reads links between notes as plain words, and every answer links to the note it came from. Click the number and you see the passage; a cited note opens in Obsidian.

With Apple Intelligence or a model it downloads, it works with Wi-Fi off and nothing leaves your Mac; it can also use your Ollama or LM Studio models (Pro plans). Obsidian vaults are included on every plan. The trade-offs: it's Mac only (Apple silicon, macOS 15 or later), it doesn't write into your notes, and it doesn't run inside Obsidian or on your phone. It's free for 30 days, then from $9.99 a month.

Questions

Sources, checked 9 October 2026: Obsidian changelog, Obsidian manifesto, Obsidian pricing, Web Clipper Interpreter, Obsidian CLI, plugin security, Obsidian plugin statistics, Copilot and its local model guide, Copilot pricing, Smart Connections, Smart Connections Pro, Smart Composer, Local GPT, Smart Second Brain, Khoj for Obsidian. Citenook's screenshot is from the app with a made-up vault.

Related: LM Studio vs Ollama · Chat with 1,000 PDFs offline on a Mac · Citenook for researchers