Obsidian MCP
Read, write, search, and navigate your Obsidian notes using natural language
obsidian-mcp
An MCP (Model Context Protocol) Server that gives AI assistants direct access to your Obsidian vault. Read, write, search, and navigate notes using natural language — with any MCP-compatible client.
Contents
Prerequisites
- Node.js 18+ — nodejs.org
- An Obsidian vault (a folder of
.mdfiles — no Obsidian app required at runtime)
Installation
You don't run the server manually — your AI client (Claude Desktop, Cursor, etc.) launches it automatically when it starts. All you need to do is build the project once and point your client's config at the output file.
1. Clone and build:
git clone <repo-url> obsidian-mcp
cd obsidian-mcp
npm install
npm run build
This produces dist/index.js — the file every client config will reference.
2. Find your vault path. This is the folder Obsidian opens as your vault, e.g. /Users/yourname/Documents/MyVault.
3. Follow the setup for your client below. Each config tells the client:
- where the built file is (
dist/index.js) - what vault to use (
OBSIDIAN_VAULT_PATH)
Client Setup
All clients use stdio transport — the server runs as a local subprocess on your machine. No hosting required.
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"obsidian": {
"command": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
}
}
Quit and relaunch Claude Desktop. A hammer icon (🔨) in the chat input confirms the server is connected.
Claude Code (CLI)
Register the server with the claude mcp add command:
claude mcp add obsidian \
node /absolute/path/to/obsidian-mcp/dist/index.js \
-e OBSIDIAN_VAULT_PATH=/path/to/your/vault
Verify it's registered:
claude mcp list
The tools are now available in any Claude Code session.
Cursor
Open Settings → Cursor Settings → MCP (or edit ~/.cursor/mcp.json):
{
"mcpServers": {
"obsidian": {
"command": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
}
}
Restart Cursor. The tools appear automatically in Cursor's AI chat and Composer.
VS Code
VS Code supports MCP servers through several extensions. Configuration goes in .vscode/mcp.json (project-scoped) or your user settings.json (global).
GitHub Copilot (VS Code 1.99+)
Create .vscode/mcp.json in your project, or add to User Settings (JSON):
{
"servers": {
"obsidian": {
"type": "stdio",
"command": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
}
}
In GitHub Copilot Chat, switch to Agent mode (@workspace) — the obsidian tools will be available automatically.
Cline
Open Cline's settings panel → MCP Servers → Add Server → paste:
{
"obsidian": {
"command": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
}
Continue
Edit ~/.continue/config.json:
{
"mcpServers": [
{
"name": "obsidian",
"command": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
]
}
Zed
Edit ~/.config/zed/settings.json:
{
"context_servers": {
"obsidian": {
"command": {
"path": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
}
}
}
The tools are available in Zed's AI Assistant panel.
Ollama (via mcphost)
Ollama doesn't support MCP natively. Use mcphost as a bridge — it wraps any MCP server and connects it to a local Ollama model.
1. Install mcphost:
go install github.com/mark3labs/mcphost@latest
2. Create a config file (~/.mcphost/config.json):
{
"mcpServers": {
"obsidian": {
"command": "node",
"args": ["/absolute/path/to/obsidian-mcp/dist/index.js"],
"env": {
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"
}
}
}
}
3. Start a chat session with any Ollama model:
mcphost --model ollama:qwen2.5:14b
Tool-use quality depends heavily on the model. Recommended:
qwen2.5:14b,llama3.1:8b,mistral-nemo. Models need to support function/tool calling to use MCP tools reliably.
Available Tools
| Tool | Description |
|---|---|
obsidian_list_notes | List notes in the vault, optionally filtered to a folder. Paginated. |
obsidian_read_note | Read the full content of a note by vault-relative path. |
obsidian_create_note | Create a new note with optional YAML frontmatter. |
obsidian_update_note | Overwrite an existing note's content and frontmatter. |
obsidian_append_to_note | Append content to a note (creates it if missing). |
obsidian_delete_note | Permanently delete a note. |
obsidian_move_note | Move or rename a note to a new path. |
obsidian_get_note_metadata | Read only the frontmatter and tags, without loading the body. |
obsidian_search_notes | Full-text search across all notes (case-insensitive). |
obsidian_search_by_tag | Find all notes with a specific #tag. |
obsidian_get_backlinks | Find all notes that [[link]] to a given note. |
obsidian_list_folders | List folders in the vault with note counts. |
obsidian_create_folder | Create a new folder (parents created automatically). |
All tools accept a response_format parameter: "markdown" (default, human-readable) or "json" (structured, for programmatic use).
Example prompts
"Summarise everything in my projects folder"
"Create a note called 'Meeting Notes 2025-04-11' with today's agenda"
"Find all notes tagged #todo and list what's incomplete"
"What notes link back to my 'Home' note?"
"Search for anything mentioning the Q2 launch"
"Append '- [ ] Follow up with design team' to my Daily Note"
Security
- Path traversal protection — all paths are validated to stay within
OBSIDIAN_VAULT_PATH - Symlink protection — symlinks inside the vault that point outside it are blocked
- File size limit — notes larger than 5 MB are rejected to prevent memory exhaustion
- Response size limit — responses are truncated at 25,000 characters with a clear notice
- No network access — the server reads and writes local files only; no outbound requests
Extending
The codebase is modular by design. To add a new set of tools:
- Create
src/tools/my-feature.tsand export aregisterMyFeatureTools(server, vault)function - Add it to
src/tools/index.ts:
import { registerMyFeatureTools } from './my-feature.js';
export function registerAllTools(server: McpServer, vault: VaultService): void {
registerNoteTools(server, vault);
registerSearchTools(server, vault);
registerFolderTools(server, vault);
registerMyFeatureTools(server, vault); // ← add this
}
- Run
npm run build
To add vault capabilities (e.g. reading canvas files, template expansion), extend VaultService in src/services/vault.ts and call the new methods from your tool handlers.
Máy chủ liên quan
Kone.vc
nhà tài trợMonetize your AI agent with contextual product recommendations
XenonFlare MCP Server
his server allows AI assistants (like Claude) to manage your social media content.
FORMLOVA
Chat-first form service with 127 MCP tools. Create, manage, and analyze forms through conversation.
JIRA
Integrate Atlassian JIRA into any MCP-compatible application to manage issues and projects.
HubSpot
Interact with the HubSpot CRM API to manage contacts, companies, and deals.
AI2Fin | Intelligent Financial Management
Tools for expense tracking, tax deductions, bill management, and financial analysis.
Shippy MCP
Ship work. Earn royalties.
stravacz-mcp
Order meals with strava.cz
Hyperpost
An AI-native publishing engine for persona-driven content creation and multi-platform publishing.
Awesome Agent Skills MCP Server
A Model Context Protocol (MCP) server that provides access to 100+ curated AI agent skills from the VoltAgent Awesome Agent Skills collection.
Prompeteer
Generate expert-level AI prompts for 140+ platforms, score quality with 16-dimension Prompt Score analysis, and manage prompts in PromptDrive library