Obsidian MCP
Model Context Protocol server that lets Claude read, write, and reorganize your Obsidian vault.
Obsidian power users on Claude Desktop (or another MCP client) who want an AI assistant to draft, refactor, search, and re-tag Markdown notes directly inside a local vault.
People who need semantic vault search, Dataview/plugin integration, mobile access, or who cannot commit to backing up their vault before granting write access.
Obsidian MCP is a lightweight, open-source Model Context Protocol server by Steven Stavrakis that exposes an Obsidian vault to any MCP-compatible AI client, most notably Claude Desktop. Instead of pasting notes into a chat, you point the server at one or more vault directories and the assistant gains a set of typed tools it can call directly against your Markdown files: read a note, create a new note, edit or delete an existing one, move or rename notes, create directories, run vault-wide searches, and add, remove, or rename tags. A separate listing tool advertises available vaults so multi-vault users can scope requests. Under the hood it is a small TypeScript/Node.js process (Node 20+) that runs locally over stdio, so your notes never leave the machine and no cloud service is involved. Installation is a one-line entry in Claude Desktop's `claude_desktop_config.json` pointing to `npx obsidian-mcp` with the vault path, or a single Smithery CLI command. It is best suited to Obsidian power users who already live in Claude Desktop and want the assistant to draft new notes into the right folder, refactor tag taxonomies, summarize existing notes, or find and stitch together related material without leaving the chat. Because it has full write access, the maintainer strongly recommends backing the vault up (or committing to git) before letting an agent loose. It is deliberately narrow in scope — no plugin API, no Dataview queries, no canvas/graph manipulation, no live sync of Obsidian's own state — but for pure Markdown + tag workflows it covers the operations most people actually want.
The most straightforward way to give Claude real hands on your Obsidian vault. It is small, honest about its limits, and does the boring CRUD-plus-tags work well — which is exactly what an MCP server should do. Just treat the 'back up your vault first' warning as non-negotiable and keep the vault under git.
— The AI Tool Bible editorial team
Pros
- ✅ Twelve concrete note and tag tools cover the everyday read/write/search operations most Obsidian users need.
- ✅ Runs entirely locally over stdio — vault contents never leave the machine and no API keys or accounts are required.
- ✅ Trivial install: one npx command in Claude Desktop config, or a single Smithery CLI line.
- ✅ Multi-vault aware — expose several vaults and let the assistant list and target them explicitly.
- ✅ MIT-licensed TypeScript source that is small enough to audit or fork before pointing an agent at your notes.
- ✅ Works with any MCP client, not just Claude Desktop, so the same server plugs into Cline, Cursor, Zed, or custom agents.
Cons
- ⚠️ Full write access with no built-in undo — a bad agent call can rename, move, or delete notes irreversibly, so backups or git are mandatory.
- ⚠️ Does not talk to the running Obsidian app or its plugin API: no Dataview queries, no canvas, no graph view, no plugin actions.
- ⚠️ Search is plain text over Markdown, not the semantic/embedding search users may expect from an 'AI' Obsidian integration.
- ⚠️ Solo-maintained hobby project — the README itself notes tools are 'tested but not exhaustively' and development is ongoing.
- ⚠️ No first-party support for the Obsidian mobile app or Obsidian Sync-specific features.
- ⚠️ Requires Node.js 20+ locally, which is friction for non-technical Obsidian users used to installing community plugins from inside the app.
Use cases
Explore related
Compare with similar tools
All in MCP Servers →Airtable MCP Server
Model Context Protocol server that gives LLMs full read/write access to your Airtable bases.
Apple Notes MCP
Let Claude read your local Apple Notes over the Model Context Protocol.
AWS MCP Servers
Official AWS Labs collection of Model Context Protocol servers for connecting AI coding assistants and agents to AWS services and documentation.
Azure MCP Server
Official Microsoft MCP server that exposes 40+ Azure services to AI agents under the developer's own Entra ID credentials.
Blender MCP
Give Claude hands inside Blender — an MCP server for natural-language 3D modeling.
Brave Search MCP
Give any MCP client web and local search powered by the Brave Search API.