The XDA writer this week described handing Claude read access to their Obsidian vault and finding connections between notes they’d forgotten they made. The technique works. The workflow, uploading whole folders and pasting excerpts, does not scale. If you’re going to point an AI at years of notes, run it against the vault properly. These are the eight desktop apps we install on Windows, macOS, and Linux to make Obsidian vaults searchable, summarisable, and interrogable without leaving the note environment.
What to look for in an AI vault analysis app
- Local vs cloud. Sending vault contents to a hosted model is fine for public notes. For private ones, prefer a local model or an app that lets you self-host the embedding.
- Retrieval quality. Vector search on chunks of notes matters more than model size. A tight embedding pipeline over 500 notes beats a huge model with poor context selection.
- Where the answer lives. Some tools inject AI into Obsidian’s sidebar; others run a separate app that watches the vault. Both work, they have different failure modes.
- Continuous indexing. Notes change constantly. A tool that reindexes on save is worth twice one that indexes once a week.
- Source citations. An answer without the note it came from is a summary you can’t verify. Insist on links back.
- Cost. Cloud tokens add up on large vaults. Local models are free per query but need a capable GPU.
Quick comparison
| App | Best for | Local model | Free tier | Cost after |
|---|---|---|---|---|
| Obsidian | The vault itself | N/A | Yes | $8/month Sync |
| Reor | Local AI vault built from scratch | Yes | Free (open source) | None |
| Smart Connections plugin | In-vault semantic search | Yes | Free | Optional Supporter |
| Copilot for Obsidian | Multi-model chat sidebar | Yes | Free | BYO API key |
| Text Generator plugin | Templated AI writes into notes | Yes | Free | BYO API key |
| AnythingLLM | Full RAG app around any folder | Yes | Free desktop | Cloud plan starts free |
| Cursor | Coder-focused AI on any folder | No | Free tier | $20/month Pro |
| NotebookLM | Cloud grounded chat over uploaded notes | No | Free | Free |
The 8 best AI Obsidian vault analysis apps
1. Obsidian, best for the vault itself
Obsidian is the note environment the rest of this list sits on top of. The 2026 build shipped Obsidian Bases and a properties overhaul that make structured queries against note metadata far cleaner than before. It ships no AI of its own, and that’s the correct choice: plugins and companion apps compete for the same slot.
Where it falls short: No built-in AI. Sync costs extra. Mobile app is capable but a step behind desktop.
Pricing:
- Free: personal use, all core features
- Paid: Sync from $8/month, Publish from $10/month
Platforms: Windows, macOS, Linux, iOS, Android.
Download: Obsidian
Bottom line: The vault. Pick your AI on top from the seven below.
2. Reor, best for a local AI vault built from scratch
Reor is an open-source note app that treats vector search as a first-class citizen. Every note is embedded on save with a local model, and every note automatically surfaces its most related neighbours in a sidebar. It also imports Obsidian markdown one-to-one, so migrating is a matter of pointing it at the folder.
Where it falls short: Not a plugin, it’s a separate app. Runs alongside Obsidian rather than inside it.
Pricing:
- Free: open source, AGPL
- Paid: none
Platforms: Windows, macOS, Linux.
Download: Reor
Bottom line: The best answer if you want AI-first notes without sending anything to the cloud.
3. Smart Connections plugin, best for in-vault semantic search
Smart Connections is the community plugin that puts the “notes like this one” sidebar and a chat-with-vault pane directly in Obsidian. It runs local embeddings by default and lets you bring an API key from OpenAI or Anthropic for the chat side. This is the plugin XDA’s Claude workflow should have been.
Where it falls short: First embedding pass on a big vault is slow. The chat side is only as good as the model you bring.
Pricing:
- Free: full plugin
- Paid: optional Supporter tier for early builds
Platforms: anywhere Obsidian runs.
Download: Smart Connections
Bottom line: The default install for anyone who wants AI in Obsidian and doesn’t want a second app.
4. Copilot for Obsidian, best for a multi-model chat sidebar
Copilot for Obsidian is the plugin that lets you switch between GPT, Claude, Gemini, and local models in a single sidebar. Vault chat with citations is on by default. Where Smart Connections is opinionated, Copilot is a swiss army knife.
Where it falls short: Configuration surface is huge. BYOK model means real cloud model use costs real dollars.
Pricing:
- Free: full plugin
- Paid: bring your own API keys
Platforms: anywhere Obsidian runs.
Download: Copilot for Obsidian
Bottom line: The pick for people who want to try multiple models against the same vault.
5. Text Generator plugin, best for templated AI writes into notes
Text Generator is the community plugin that lets you write a template once, then run it against any note or selection. Great for consistent summaries, weekly reviews, and inbox triage. It also supports local models via LM Studio and Ollama.
Where it falls short: Not a chat app. If you want a conversation, use Copilot or Smart Connections.
Pricing:
- Free: full plugin
- Paid: bring your own API keys
Platforms: anywhere Obsidian runs.
Download: Text Generator plugin
Bottom line: The template engine to add AI to a routine, not to a search box.
6. AnythingLLM, best for full RAG app around any folder
AnythingLLM is a desktop app that indexes any folder into workspaces you can chat with. Point it at your Obsidian vault, then a client folder, then a code repo, and switch between them in one UI. Runs Ollama, LM Studio, and cloud models interchangeably.
Where it falls short: Lives outside Obsidian. No back-links into notes; retrieval is the strength, editing is not.
Pricing:
- Free: desktop app, open source
- Paid: Cloud team plan from $50/month
Platforms: Windows, macOS, Linux.
Download: AnythingLLM
Bottom line: The pick when Obsidian is one of several corpora you want to chat with.
7. Cursor, best for coder-focused AI on any folder
Cursor is the AI-first code editor, and it treats an Obsidian vault as a workspace like any other folder. Point it at the vault, ask “which notes discuss X?” and it does grep-plus-embeddings across every markdown file. Not what most note-takers want, but the developer crossover is real.
Where it falls short: No note-writer conveniences. Every file is treated as code.
Pricing:
- Free: hobby tier
- Paid: Pro at $20/month
Platforms: Windows, macOS, Linux.
Download: Cursor
Bottom line: The pick when your vault is really “notes plus a lot of code.”
8. NotebookLM, best for cloud grounded chat over uploaded notes
NotebookLM is Google’s grounded chat product. Upload a set of markdown files and ask questions; the answers cite the specific note passages they came from. The 2026 build supports 300-source notebooks, so a mid-sized vault fits in a single project.
Where it falls short: Upload-based, not synced. The vault has to be re-uploaded when it changes. Cloud-only.
Pricing:
- Free: full features, generous limits
- Paid: NotebookLM Plus with Google One AI Premium
Platforms: web on Windows, macOS, Linux, ChromeOS.
Download: NotebookLM
Bottom line: Try it first for a one-off “read all my notes about X” project before wiring up a persistent tool.
How to pick the right one
- If you want AI inside Obsidian and nothing else: Smart Connections.
- If you want to try more than one model: Copilot for Obsidian.
- If you want everything local: Reor.
- If your vault is one of several corpora you’d chat with: AnythingLLM.
- If you already live in a code editor: Cursor.
- If you just want to answer one question this weekend: NotebookLM.
FAQ
Is it safe to send my notes to a cloud AI? Only if you’re OK with the vendor’s data policy. For anything sensitive, use Reor, Smart Connections with a local model, or AnythingLLM with Ollama.
Which local model works well for note retrieval? Any embedding model works. For chat, small quantised models like Llama 3 8B or Mistral 7B are enough for note summarisation on a laptop with a modern GPU.
Do these tools support graph-based search? Smart Connections and Reor both surface semantic neighbours, which is the graph most users actually want. Obsidian’s built-in link graph is a separate thing.
Can I keep using Obsidian Sync with these plugins? Yes. Sync moves the raw markdown; the embeddings live on each device.
Does any of this work with an encrypted vault? Only if you decrypt first. AI plugins can only read the plain-text markdown Obsidian sees.