# Obsidian MCP [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/seynurmammad0v/obsidian-mcp  
**GitHub Stars:** 1  
**npm Downloads (last month):** 17098  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/obsidian-mcp

## Description
Turns Obsidian vaults into AI-queryable knowledge graphs with 12 tools

## Tools
Capabilities this server exposes over MCP:

- **vault_stats** — Overview: total notes, tags, links, orphans, components, top tags/hubs
- **read_note** — Full content + frontmatter + links + backlinks + sections
- **search_notes** — Filter by tags, folder, frontmatter, keyword. All combinable
- **graph_neighbors** — Direct links + backlinks for a note
- **graph_traverse** — BFS/DFS from a note, N hops deep with tag filtering
- **graph_shortest_path** — Shortest link chain between two notes
- **graph_analyze** — Components, orphans, bridges, hubs, dead links
- **create_note** — Create with content + frontmatter
- **patch_note** — Append, prepend, or replace a section. Update frontmatter
- **move_note** — Rename/move + auto-update all backlinks across vault
- **delete_note** — Delete + report broken links
- **manage_tags** — Add/remove/rename tags (vault-wide rename supported)

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "obsidian-mcp": {
    "command": "npx",
    "args": ["-y","obsidian-mcp"]
  }
}
```

## Documentation

## What the Obsidian MCP MCP server does

Obsidian MCP MCP server exposes an Obsidian vault to MCP-compatible AI clients as a searchable knowledge graph. It reads `.md` files, parses note content and frontmatter, and follows Obsidian `[[wikilinks]]` to represent relationships between notes. The vault does not need a particular folder layout or metadata scheme.

Its 12 tools cover discovery, graph analysis, and file manipulation. Agents can inspect vault-wide statistics, retrieve a note with its links and backlinks, search using combined filters, and examine graph structure such as connected components, orphans, bridges, hubs, and dead links. They can also create, edit, move, delete, and tag notes.

## How it works

At startup, the process scans the vault’s Markdown files and constructs the graph in memory. The README reports an approximate startup time of one to three seconds for a vault containing 1,000 notes. A file watcher then applies changes incrementally, with an expected update time of about five milliseconds per change.

Graph tools support direct-neighbor queries, breadth-first or depth-first traversal with a configurable hop limit, and shortest-path searches between notes. Traversal can also filter by tags. When a write tool changes the vault, the filesystem and in-memory graph are updated atomically. Moving a note also updates references to it throughout the vault, while deleting a note reports links that may now be broken.

The process excludes `.obsidian/`, `.trash/`, and `node_modules/` by default. Additional paths can be ignored through a configuration file in the vault root.

## Setup and configuration

Run the server with Node’s package runner and provide the vault path:

```bash
npx obsidian-mcp --vault ~/my-obsidian-vault
```

The `--vault` argument is required. Optional command-line settings enable verbose output and select an error, warning, information, or debug log level.

Claude Desktop can launch it through an MCP configuration entry using `npx -y obsidian-mcp --vault /path/to/vault`. Claude Code can register the same command with `claude mcp add`.

An optional `.obsidian-mcp.json` file can set ignored paths, the file-watch debounce interval, maximum graph traversal depth, and maximum search results. The documented defaults are `100` for `watchDebounce`, `10` for `maxTraversalDepth`, and `500` for `maxSearchResults`.

## Tools and capabilities

Reading and discovery tools provide vault statistics, complete note data, and combined searches across keywords, folders, tags, and frontmatter. Graph tools expose backlinks and outgoing links, multi-hop traversal, shortest link chains, and structural analysis.

Writing tools create notes with content and frontmatter, patch note sections or frontmatter, move and rename files, remove notes, and add, remove, or rename tags. Tag renaming can apply across the vault. The Obsidian MCP MCP server therefore fits workflows that need both knowledge retrieval and controlled vault maintenance.

## Limitations and notes

The vault must be supplied as a local path, and the server relies on Markdown files and Obsidian-style wikilinks for its graph. No database, running Obsidian instance, or API key is required according to the project documentation. The graph is held in memory and is initially built by scanning the vault, so large vaults may have a startup delay.

It does not describe authentication, remote vault access, or client-specific configuration for other applications. Built-in exclusions and configured ignore paths prevent selected files and directories from entering normal processing.

_Full upstream README: https://allmcps.com/mcp/obsidian-mcp/readme_