Obsidian MCP Server

Obsidian MCP Server

A Model Context Protocol (MCP) server that enables AI assistants like Claude to interact with your Obsidian vault. This server provides tools for reading, creating, searching, and managing notes in Obsidian through the Local REST API plugin.

Features

• 📖 Read & write notes - Full access to your Obsidian vault with automatic overwrite protection
• 🔍 Smart search - Find notes by content, tags, or modification date
• 📁 Browse vault - List and navigate your notes by directory
• 🏷️ Tag management - Add, remove, and organize tags in frontmatter
• 📊 Note insights - Get statistics like word count and link analysis
• 🎯 AI-optimized - Clear error messages and smart defaults for better AI interactions
• 🔒 Secure - API key authentication with local-only connections

Prerequisites

• Obsidian with the Local REST API plugin installed and enabled
• Python 3.10+ installed on your system
• Node.js (optional, for running MCP Inspector)

Installation

Quick Install

1. Install and configure Obsidian:
   • Install the Local REST API plugin in Obsidian
   • Enable the plugin in Settings > Community plugins
   • Go to Settings > Local REST API
   • Copy your API key (you'll need this for step 2)
2. Configure your AI tool:Edit your Claude Desktop config file:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
   { "mcpServers": { "obsidian": { "command": "uvx", "args": ["obsidian-mcp"], "env": { "OBSIDIAN_REST_API_KEY": "your-api-key-here" } } } }

   Add to your Cursor settings:

   • Project-specific: .cursor/mcp.json in your project directory
   • Global: ~/.cursor/mcp.json in your home directory
   { "mcpServers": { "obsidian": { "command": "uvx", "args": ["obsidian-mcp"], "env": { "OBSIDIAN_REST_API_KEY": "your-api-key-here" } } } }

   Then: Open Settings → Cursor Settings → Enable MCP

   Edit your Windsurf config file:

   • Location: ~/.codeium/windsurf/mcp_config.json
   { "mcpServers": { "obsidian": { "command": "uvx", "args": ["obsidian-mcp"], "env": { "OBSIDIAN_REST_API_KEY": "your-api-key-here" } } } }

   Then: Open Windsurf Settings → Advanced Settings → Cascade → Add Server → Refresh

   Replace your-api-key-here with the API key you copied from Obsidian.

3. Restart your AI tool to load the new configuration.

That's it! The server will now be available in your AI tool with access to your Obsidian vault.

Note: This uses uvx which automatically downloads and runs the server in an isolated environment. Most users won't need to install anything else. If you don't have uv installed, you can also use pipx install obsidian-mcp and change the command to "obsidian-mcp" in the config.

Try It Out

Here are some example prompts to get started:

• "Show me all notes I modified this week"
• "Create a new daily note for today with my meeting agenda"
• "Search for all notes about project planning"
• "Read my Ideas/startup.md note"

Development Installation

1. Clone the repository:
   git clone https://github.com/natestrong/obsidian-mcp cd obsidian-mcp
2. Set up Python environment:
   # Using pyenv (recommended) pyenv virtualenv 3.12.9 obsidian-mcp pyenv activate obsidian-mcp # Or using venv python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate
3. Install dependencies:
   pip install -r requirements.txt
4. Set up Obsidian Local REST API:
   • Install the Local REST API plugin in Obsidian
   • Enable the plugin in Obsidian settings
   • Copy the API key from the plugin settings
   • Note the port number (default: 27124)
5. Configure environment variables:
   export OBSIDIAN_REST_API_KEY="your-api-key-here" export OBSIDIAN_API_URL="https://localhost:27124" # if not using default
6. Add to Claude Desktop (for development):Edit your Claude Desktop config file:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
   { "mcpServers": { "obsidian": { "command": "/path/to/python", "args": ["-m", "src.server"], "cwd": "/path/to/obsidian-mcp", "env": { "PYTHONPATH": "/path/to/obsidian-mcp", "OBSIDIAN_REST_API_KEY": "your-api-key-here", "OBSIDIAN_API_URL": "https://localhost:27124" } } } }

Project Structure

Available Tools

Note Management

read_note

Read the content and metadata of a specific note.

Parameters:

• path: Path to the note (e.g., "Daily/2024-01-15.md")

Returns:

create_note

Create a new note or update an existing one.

Parameters:

• path: Path where the note should be created
• content: Markdown content of the note
• overwrite (default: false): Whether to overwrite existing notes

update_note

Update the content of an existing note.

⚠️ IMPORTANT: By default, this tool REPLACES the entire note content. Always read the note first if you need to preserve existing content.

Parameters:

• path: Path to the note to update
• content: New markdown content (REPLACES existing content unless using append)
• create_if_not_exists (default: false): Create if doesn't exist
• merge_strategy (default: "replace"): How to handle content
  • "replace": Overwrites entire note content (default)
  • "append": Adds new content to the end of existing content

Safe Update Pattern:

delete_note

Delete a note from the vault.

Parameters:

• path: Path to the note to delete

Search and Discovery

search_notes

Search for notes containing specific text.

Parameters:

• query: Search query (supports Obsidian search syntax)
• context_length (default: 100): Number of characters to show around matches

Note: Search functionality may have connectivity issues with some REST API configurations.

search_by_date

Search for notes by creation or modification date.

Parameters:

• date_type (default: "modified"): Either "created" or "modified"
• days_ago (default: 7): Number of days to look back
• operator (default: "within"): Either "within" (last N days) or "exactly" (exactly N days ago)

Returns:

Example usage:

• "Show me all notes modified this week" → search_by_date("modified", 7, "within")
• "Find notes created in the last 30 days" → search_by_date("created", 30, "within")
• "What notes were modified exactly 2 days ago?" → search_by_date("modified", 2, "exactly")

list_notes

List notes in your vault with optional recursive traversal.

Parameters:

• directory (optional): Specific directory to list (e.g., "Daily", "Projects")
• recursive (default: true): List all notes recursively

Returns:

Organization

move_note

Move a note to a new location.

Parameters:

• source_path: Current path of the note
• destination_path: New path for the note
• update_links (default: true): Update links in other notes (future enhancement)

add_tags

Add tags to a note's frontmatter.

Parameters:

• path: Path to the note
• tags: List of tags to add (without # prefix)

remove_tags

Remove tags from a note's frontmatter.

Parameters:

• path: Path to the note
• tags: List of tags to remove

get_note_info

Get metadata and statistics about a note without retrieving its full content.

Parameters:

• path: Path to the note

Returns:

Testing

Running Tests

Testing with MCP Inspector

1. Ensure Obsidian is running with the Local REST API plugin enabled
2. Run the MCP Inspector:
   npx @modelcontextprotocol/inspector python -m src.server
3. Open the Inspector UI at http://localhost:5173
4. Test the tools interactively with your actual vault

Integration with Claude Desktop

For development installations, see the Development Installation section above.

Enhanced Error Handling

The server provides detailed, actionable error messages to help AI systems recover from errors:

Example Error Messages

Invalid Path:

Empty Search Query:

Invalid Date Parameters:

Troubleshooting

"Connection refused" error

• Ensure Obsidian is running
• Verify the Local REST API plugin is enabled
• Check that the port matches (default: 27124)
• Confirm the API key is correct
• The enhanced error will show the exact URL and port being used

"Certificate verify failed" error

• This is expected with the Local REST API's self-signed certificate
• The server handles this automatically

"Module not found" error

• Ensure your virtual environment is activated
• Run from the project root: python -m src.server
• Verify PYTHONPATH includes the project directory

Empty results when listing notes

• Specify a directory when using list_notes (e.g., "Daily", "Projects")
• Root directory listing requires recursive implementation
• Check if notes are in subdirectories

Tags not updating

• Ensure notes have YAML frontmatter section
• Frontmatter must include a tags: field (even if empty)

Best Practices for AI Assistants

Preventing Data Loss

1. Always read before updating: The update_note tool REPLACES content by default
2. Use append mode for additions: When adding to existing notes, use merge_strategy="append"
3. Check note existence: Use read_note to verify a note exists before modifying
4. Be explicit about overwrites: Only use overwrite=true when intentionally replacing content

Recommended Workflows

Safe note editing:

1. Read the existing note first
2. Modify the content as needed
3. Update with the complete new content

Adding to daily notes:

• Use merge_strategy="append" to add entries without losing existing content

Creating new notes:

• Use create_note with overwrite=false (default) to prevent accidental overwrites

Security Considerations

• Keep your API key secret - never commit it to version control
• The server validates all paths to prevent directory traversal attacks
• All communication with Obsidian uses HTTPS (self-signed certificate)
• The server only accepts local connections through the REST API

Development

Code Style

• Uses FastMCP framework for MCP implementation
• Pydantic models for type safety and validation
• Modular architecture with separated concerns
• Comprehensive error handling and user-friendly messages

Adding New Tools

1. Create tool function in appropriate module under src/tools/
2. Add Pydantic models if needed in src/models/
3. Register the tool in src/server.py with the @mcp.tool() decorator
4. Include comprehensive docstrings
5. Add tests in tests/
6. Test with MCP Inspector before deploying

Publishing (for maintainers)

To publish a new version to PyPI:

Users can then install and run with:

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-tool)
3. Write tests for new functionality
4. Ensure all tests pass
5. Commit your changes (git commit -m 'Add amazing tool')
6. Push to the branch (git push origin feature/amazing-tool)
7. Open a Pull Request

License

MIT License - see LICENSE file for details

Acknowledgments

• Anthropic for creating the Model Context Protocol
• Obsidian team for the amazing note-taking app
• coddingtonbear for the Local REST API plugin
• dsp-ant for the FastMCP framework