Obsidian MCP Second Brain Server

An MCP (Model Context Protocol) server that provides intelligent read-only access to your Obsidian vault, enabling it to function as a "second brain" for LLMs.

typescript

4

GitHub Stars

typescript

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "comfucios-obsidian-mcp-sb": {
      "command": "npx",
      "args": [
        "-y",
        "@comfucios/obsidian-mcp-sb",
        "--vault-path",
        "/path/to/your/vault"
      ]
    }
  }
}

Obsidian MCP Second Brain Server provides a read-only MCP endpoint for querying and retrieving content from your Obsidian vault. It enables semantic search, metadata filters, and smart summaries while keeping your vault safe from edits during AI interactions.

How to use

You run the Obsidian MCP Second Brain Server as a local MCP client, then connect your preferred MCP consumer (such as Claude Code, Claude Desktop, or another MCP client) to this server. The server is read-only and lets you search notes, filter by path and tags, retrieve full note content, generate summaries of a collection, and access recent notes without modifying any files.

How to install

Prerequisites you need before starting:

Install Node.js and npm. Ensure you have a modern Node.js environment available on your system.

Manual (no full installation required to run): use the following command to start the server directly with your vault path.

npx -y @comfucios/obsidian-mcp-sb --vault-path "/path/to/your/vault"

Local development setup

If you are developing or testing locally, install dependencies, build the project, and link the package globally so you can run it easily from your shell.

cd obsidian-mcp-sb
npm install
npm run build
npm link

Configuration & usage notes

The server is designed to be read-only. It supports search, retrieval, and summarization, but does not create or modify notes.
To point the server at your vault, pass the vault path as shown in the start command: --vault-path "/path/to/your/vault".

Claude integration guidance

If you use Claude Code or Claude Desktop, configure the MCP server to point to the local npx command with the vault path, as described in the next section.

Security and storage considerations

Memory mode is available for small vaults to speed up indexing during development.
The server uses path sanitization and input validation to protect against path traversal and overly large inputs.

Troubleshooting & tips

If you encounter slow searches on very large vaults, ensure you are using the SQLite-based storage mode by default, which provides persistent indexing.
For quick tests, the in-memory mode can be faster for very small vaults, but it requires the --use-memory option when starting in a development context.

Storage and architecture overview

Default storage uses SQLite-based indexing with FTS5 for efficient full-text search and persistent caching.
In-memory storage uses Fuse.js for fuzzy search and is intended for small vaults or development.

Notes on features

Semantic search across notes with fuzzy matching.
Tag-based and path-based filtering to narrow results.
Temporal queries by creation or modification dates.
Metadata filtering and full note retrieval.
Smart summaries and statistics for note collections.

Example usage scenarios

Quick retrieval of notes about a specific topic, filtered by tag like work/puppet or tech/golang.
Generating a concise summary of notes related to a project or topic to share with an AI assistant.

Notes on deployment options

Use the stdio configuration for local runs with npm/yarn-driven workflows.