Obsidian MCP Plugin

📋 Community Plugin Status: This plugin is currently under review for inclusion in the Obsidian Community Plugins directory. Submitted on July 10, 2025 - PR #6991

A high-performance Model Context Protocol (MCP) server implemented as an Obsidian plugin, providing AI tools with direct vault access through HTTP transport.

Overview

This plugin brings MCP capabilities directly into Obsidian, eliminating the need for external servers or the REST API plugin. It provides semantic, AI-optimized operations that consolidate multiple tools into intelligent workflows with contextual hints.

What are "Semantic Tools"?

Unlike basic tools that just return data, semantic tools are self-guiding. When an AI agent uses a semantic tool, it doesn't just get the requested information—it also receives smart suggestions about what to do next. This creates a natural workflow where each action leads intelligently to the next, helping AI agents complete complex tasks without getting stuck or needing constant human guidance.

Example: When searching for a file, a semantic tool doesn't just return search results. It also suggests: "Found 3 files matching 'meeting notes'. Consider using view to read the most recent one, or graph traverse to explore related documents."

Key Features

• Direct Obsidian Integration: Runs natively within Obsidian for maximum performance
• HTTP MCP Transport: Compatible with Claude Desktop, Claude Code, Cline, and other MCP clients
• Semantic Operations: Enhanced search with Obsidian operators, intelligent fragment retrieval, and workflow guidance
• No External Dependencies: No need for the REST API plugin or external servers
• High Performance: Sub-100ms response times with direct vault access
• Concurrent Sessions: Support for multiple AI agents working simultaneously (v0.5.8+)
• Worker Thread Processing: CPU-intensive operations run in parallel threads for non-blocking performance
• Path Exclusions: .mcpignore file support for blocking sensitive files from AI access
• Cloud Sync Friendly: Smart retry logic handles file sync delays from iCloud, OneDrive, Dropbox, and other services

🔐 Path Exclusions with .mcpignore

Protect sensitive files and directories from AI access using .gitignore-style patterns.

Overview

The plugin supports a .mcpignore file in your vault root that allows you to exclude specific files and directories from MCP operations. This provides a security layer ensuring that AI tools cannot access your private or sensitive content.

Setup

1. Enable in Settings: Toggle "Enable Path Exclusions (.mcpignore)" in plugin settings
2. Create Template: Click "Create Template" to generate a starter .mcpignore file with examples
3. Edit Patterns: Use any text editor to add exclusion patterns

Pattern Syntax

The .mcpignore file uses the same syntax as .gitignore:

# Directories
private/              # Excludes 'private' directory and ALL its contents
/private/            # Only excludes 'private' at vault root
work/*/confidential/ # Excludes 'confidential' dirs one level under work/

# Files
secrets.md           # Excludes this specific file
*.private           # All files ending with .private
daily/*.md          # All .md files directly in daily/

# Negation (whitelist)
!public.private     # Allow this specific file despite *.private rule

Features

• Auto-reload: Changes to .mcpignore take effect immediately
• Context Menu: Right-click any file/folder and select "Add to .mcpignore"
• UI Controls: Manage from plugin settings with quick access buttons
• Security: Blocked paths return PATH_BLOCKED errors with no data leakage
• Protected: The .mcpignore file itself cannot be accessed via MCP

Example .mcpignore

# Personal content
journal/
diary/
private/

# Work files
work/confidential/
clients/*/contracts/

# Temporary files
*.tmp
*.backup
.#*

# Allow specific files
!work/public-docs/

☁️ Cloud Sync Friendly

The plugin gracefully handles sync delays and file locking from cloud storage services.

Supported Services

Works automatically with:

• iCloud Drive - Handles .icloud sync artifacts and timing conflicts
• OneDrive - Manages file locks and sync delays
• Dropbox - Handles sync markers and temporary files
• Google Drive - Manages sync state transitions
• Any sync service - Universal retry logic works with all platforms

How It Works

The plugin automatically:

1. Detects sync conflicts - Recognizes EEXIST, EBUSY, and "file already exists" errors
2. Retries intelligently - Uses exponential backoff (500ms → 1s → 2s)
3. Resolves timing issues - Allows sync services time to complete operations
4. Works transparently - No configuration needed, just works

Benefits

• No phantom files - Resolves "file already exists" errors for non-existent files
• Reliable operations - File creation, updates, and deletes work consistently
• Zero configuration - Automatic detection and handling
• Cross-platform - Same behavior on macOS, Windows, and Linux

Installation

Note

: This plugin is currently pending review for the Obsidian Community Plugins directory. Until approved, please use the BRAT installation method below.

Current Installation: Via BRAT (Beta Reviewer's Auto-update Tool)

Since this plugin is not yet in the Community Plugins directory, you'll need to use BRAT to install it:

1. Install BRAT:

   • Open Obsidian Settings → Community Plugins
   • Browse and search for "BRAT"
   • Install and enable the BRAT plugin

2. Add This Plugin:

   • In Obsidian settings, go to BRAT settings
   • Click "Add Beta Plugin"
   • Enter: https://github.com/aaronsb/obsidian-mcp-plugin
   • Click "Add Plugin"

3. Enable the Plugin:

   • Go to Settings → Community Plugins
   • Find "Obsidian MCP Plugin" and enable it
   • The plugin will auto-update through BRAT

Future Installation: Via Community Plugins (After Approval)

Once this plugin is approved and available in the Obsidian Community Plugins directory:

1. Open Obsidian Settings → Community Plugins
2. Click "Browse" and search for "MCP"
3. Find "Obsidian MCP Plugin" by Aaron Bockelie
4. Click "Install" then "Enable"

For BRAT Users: After the plugin is approved, you can remove it from BRAT and install it normally through Community Plugins to receive standard updates.

Configuration

1. Enable the Server:

   • Go to plugin settings
   • Toggle "Enable HTTP Server"
   • Default port is 3001 (configurable)

2. Connect Your MCP Client:

   Claude Code

   claude mcp add obsidian http://localhost:3001/mcp --transport http
   

   Claude Desktop / Other Clients

   Option 1: Direct HTTP Transport (if your client supports it)

   {
     "mcpServers": {
       "obsidian": {
         "transport": {
           "type": "http",
           "url": "http://localhost:3001/mcp"
         }
       }
     }
   }
   

   Option 2: Via mcp-remote (recommended for Claude Desktop)

   Since Claude Desktop might not support streaming HTTP transport yet, use mcp-remote:

   {
     "mcpServers": {
       "obsidian-vault-name": {
         "command": "npx",
         "args": [
           "mcp-remote",
           "http://localhost:3001/mcp"
         ]
       }
     }
   }
   

   Configuration file locations:

   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
   • Linux: ~/.config/Claude/claude_desktop_config.json

Concurrent Sessions for Agent Swarms (v0.5.8+)

Enable multiple AI agents to work with your vault simultaneously without blocking each other:

1. Enable in Settings:

   • Go to plugin settings
   • Find "Concurrent Sessions" section
   • Toggle "Enable Concurrent Sessions for Agent Swarms"
   • Adjust max connections if needed (default: 32)

2. Architecture:

   • Each AI session gets its own isolated MCP server instance
   • True parallel processing - no blocking between sessions
   • Automatic session management with 1-hour timeout
   • Session reuse for reconnecting clients

3. Monitor Active Sessions:

   • Use the obsidian://session-info resource to view:
     • Active sessions with "This is you!" indicator
     • Session age, idle time, and request counts
     • Server pool utilization statistics

4. Use Cases:

   • Multiple Claude instances working on different parts of your vault
   • AI agent teams collaborating on research or writing projects
   • Parallel processing of large knowledge bases
   • Non-blocking operation for time-sensitive workflows

Available Tools

🗂️ vault - File and Folder Operations

• list - List files and directories
• read - Read file content with fragments for large files
• create - Create new files and directories
• update - Update existing files
• delete - Delete files and folders
• search - Enhanced search with Obsidian operators
• fragments - Get relevant excerpts from files
• move - Move files to new locations (preserves links)
• rename - Rename files in place (preserves links)
• copy - Create copies of files
• split - Split files by headings, delimiter, lines, or size
• combine - Merge multiple files with sorting options
• concatenate - Simple two-file joining

🔍 Search Operators:

• file: - Search by filename or extension (e.g., file:.png)
• path: - Search in file paths
• content: - Search only in file content
• tag: - Search for tags
• "exact phrase" - Search for exact phrases
• term1 OR term2 - Search for either term
• /regex/flags - Regular expression search

✏️ edit - Smart Editing Operations

• window - Edit with automatic content buffering
• append - Append content to files
• patch - Intelligent patching with fuzzy matching
• at_line - Edit at specific line numbers
• from_buffer - Recover content from edit buffers

👁️ view - Content Viewing and Navigation

• file - View complete files with metadata
• window - View content windows with context
• active - Get currently active file
• open_in_obsidian - Open files in Obsidian

🔄 workflow - AI Workflow Guidance

• suggest - Get contextual suggestions based on current operations

🕸️ graph - Graph Traversal and Link Analysis

• traverse - Explore connected nodes from a starting point
• neighbors - Get immediate connections of a file
• path - Find paths between two nodes
• statistics - Get link counts and statistics for files
• backlinks - Find all incoming links to a file
• forwardlinks - Find all outgoing links from a file
• search-traverse - Search-based graph traversal with snippet chains
• advanced-traverse - Multi-query traversal with strategies (breadth-first, best-first, beam-search)

Graph Features:

• links/backlinks/tags - Follow connections during traversal
• pattern/folder/tag filters - Filter by file patterns, folders, or tags
• depth/max_nodes - Control traversal depth and maximum nodes
• relevance_scores - Get relevance scores and snippet chains
• orphaned/unresolved - Support for orphaned notes and unresolved links

⚙️ system - System Operations

• info - Get vault and plugin information
• commands - List and execute Obsidian commands
• fetch_web - Fetch and convert web content to markdown

Configuration

Plugin Settings

Access plugin settings via Obsidian Settings → Community Plugins → Obsidian MCP Plugin → Settings

• HTTP Port: Port for MCP server (default: 3001)
• Enable Concurrent Sessions: Allow multiple AI agents to work simultaneously (default: enabled)
• Max Concurrent Connections: Maximum number of parallel operations (default: 32)
• Path Exclusions (.mcpignore): Enable/disable path blocking via .mcpignore file
• Enable Context Menu: Add "Add to .mcpignore" to file/folder right-click menus
• Debug Logging: Enable detailed console logging for troubleshooting

Concurrent Sessions (v0.5.8+)

The plugin supports multiple AI agents working simultaneously through session-based connection pooling:

• Each MCP client gets a unique session ID
• Sessions are isolated and tracked independently
• CPU-intensive operations (search, graph traversal) can run in parallel
• Worker threads prevent blocking the main Obsidian UI
• Sessions automatically expire after 1 hour of inactivity

Performance with Concurrent Sessions:

• Up to 32 simultaneous operations (configurable)
• Worker threads for CPU-intensive tasks
• Non-blocking UI during heavy operations
• Automatic session cleanup and resource management

MCP Resources

• obsidian://vault-info - Real-time vault metadata including file counts, active file, and plugin status
• obsidian://session-info - Active sessions and connection pool statistics (when concurrent sessions enabled)

Key Improvements Over External MCP Servers

1. Performance: Direct vault access eliminates HTTP overhead
2. Search: Uses Obsidian's native search with advanced operators
3. Integration: Access to full Obsidian API and plugin ecosystem
4. Simplicity: Single plugin installation, no external dependencies
5. Reliability: No separate processes to manage or crash

Architecture

This plugin implements the same semantic operations as obsidian-semantic-mcp but runs directly within Obsidian:

Before: AI Tool → MCP Server → REST API Plugin → Obsidian
Now:    AI Tool → MCP Plugin (within Obsidian)

The critical ObsidianAPI abstraction layer is preserved, allowing all semantic operations to work identically while gaining the performance benefits of direct integration.

Testing Status

✅ Working Features:

• All 6 semantic tools with all actions
• Enhanced search with Obsidian operators
• Graph traversal and link analysis
• Image viewing and file operations
• Fragment retrieval for large files
• Workflow hints and guidance
• Multi-vault support
• Port collision detection
• Cross-platform tested (Linux, Windows, macOS)

⚡ Performance Results:

• File operations: <10ms (vs ~50-100ms with REST API)
• Search operations: <50ms (vs ~100-300ms)
• Zero network overhead

Development

# Clone to your vault's plugins folder
git clone https://github.com/aaronsb/obsidian-mcp-plugin .obsidian/plugins/obsidian-mcp-plugin

# Install dependencies
npm install

# Build for development
npm run dev

# Build for production
npm run build

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions

License

MIT

---

This plugin brings the power of semantic MCP directly into Obsidian, providing AI tools with intelligent, high-performance vault access.