obsidian-mcp-server
MCP server that provides natural language access to an Obsidian vault, enabling reading, searching, listing, and creating notes.
README
Obsidian MCP Server
An MCP (Model Context Protocol) server that exposes an Obsidian vault as a structured interface for AI models. Connect Claude Desktop or Claude Code to your vault and interact with your notes in natural language.
What it does
Instead of manually navigating your vault, you can ask Claude:
- "Read my weekly plan for W31"
- "Search my notes for anything about HireRight"
- "What are my 5 most recently modified notes?"
- "What's the structure of my vault?"
- "Create today's daily note"
Claude calls the appropriate tool, reads or writes the file, and responds naturally.
Tools
| Tool | Description |
|---|---|
read_note(path) |
Read a note by path relative to vault root |
list_recent(n) |
List the N most recently modified notes |
search_vault(query, tag, folder) |
Full-text search with optional tag and folder filters |
get_vault_structure() |
Get the folder hierarchy and note counts |
create_note(path, content) |
Create a new note (refuses to overwrite existing) |
Architecture
Claude Desktop / Claude Code
│
│ stdio (local)
▼
MCP Server (server.py)
│
▼
Obsidian Vault (filesystem)
The server runs as a local process started by the MCP host (Claude Desktop or Claude Code). Communication happens over stdio — no network exposure, no auth required for local use.
Stack
Setup
1. Clone and install dependencies
git clone https://github.com/DusanMilic/obsidian-mcp-server.git
cd obsidian-mcp-server
uv sync
2. Set your vault path
Edit server.py and update VAULT_PATH:
VAULT_PATH = Path("/your/path/to/obsidian/vault")
3. Connect to Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"obsidian": {
"command": "/path/to/uv",
"args": [
"--directory", "/path/to/obsidian-mcp-server",
"run", "server.py"
]
}
}
}
Restart Claude Desktop. The server starts automatically.
4. Connect to Claude Code
claude mcp add --scope user obsidian -- /path/to/uv --directory /path/to/obsidian-mcp-server run server.py
Security
- All paths are validated to be within the vault root — no path traversal possible
create_noterefuses to overwrite existing notes- No write tool overwrites existing content —
create_noteis append-safe by design - Local stdio transport only — no network exposure in default configuration
Project context
Built as Phase 1 of an AI engineering learning path — the goal is to move from AI consumer to AI builder. This server is used daily to give Claude structured access to a personal Obsidian vault containing journaling, planning, and project notes.
Phase 2 will extend this into a personal finance agent. Phase 4 targets a real estate agentic analyzer over Serbian property data.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。