Obsidian CLI MCP Server
MCP server that exposes Obsidian vault operations over HTTP via the Obsidian CLI, enabling note management, search, and daily note operations.
README
Obsidian CLI MCP Server
MCP server that exposes Obsidian vault operations over HTTP. Uses the obsidian CLI via subprocess — requires the Obsidian desktop app to be running.
Placeholders: Examples use <project-root>, <vault-root>, and localhost:3100 — replace with your actual paths and host/port. All examples are generic and safe for public use.
Prerequisites
- Obsidian desktop app installed and running
- gtimeout (GNU coreutils):
brew install coreutilson macOS - Python 3.10+
Quick Start
cd <project-root> # e.g. ~/projects/obsidian-cli-mcp
pip install -r requirements.txt
cp .env.example .env # edit .env with your vault path
python -m src.server
Server listens on http://localhost:3100/mcp.
Dev mode (auto-restart on file changes):
make dev
Smoke test (run all read operations; edit hardcoded paths in scripts/smoke_test.py to match your vault):
python scripts/smoke_test.py
Tools
| Tool | Purpose |
|---|---|
health |
Check obsidian CLI and vault status |
obsidian_base_query |
Query .base database files |
obsidian_base_views |
List views in a .base file |
obsidian_read |
Read a note by name or path |
obsidian_search |
Full-text search with context |
obsidian_search_json |
Full-text search as JSON |
obsidian_daily_read |
Read today's daily note |
obsidian_daily_append |
Append to today's note (write-gated) |
obsidian_files |
List files in a folder |
obsidian_backlinks |
Notes that link to this note |
obsidian_links |
Outgoing links from a note |
obsidian_tags |
All tags with counts |
obsidian_tags_file |
Tags for a specific file |
obsidian_properties |
Frontmatter properties |
obsidian_create |
Create a note (write-gated) |
Environment Variables
| Variable | Default | Purpose |
|---|---|---|
OBSIDIAN_VAULT_PATH |
(in .env) |
Vault directory (cwd for obsidian CLI); see .env.example |
OBSIDIAN_VAULT |
(none) | Vault name when multiple vaults exist |
OBSIDIAN_TIMEOUT_SECONDS |
10 |
Timeout for obsidian CLI (base queries use 30s) |
ALLOW_WRITE_COMMANDS |
false |
Enable create, append, property:set |
MCP_HOST |
0.0.0.0 |
Bind address |
MCP_PORT |
3100 |
Server port |
MCP_TRANSPORT |
streamable-http |
stdio for Cursor command; streamable-http for URL |
Connect from Continue
Place config at <vault-root>/.continue/mcpServers/obsidian-cli-mcp.yaml:
name: Obsidian CLI
version: 1.0.0
schema: v1
mcpServers:
- name: obsidian-cli-mcp
type: streamable-http
url: http://localhost:3100/mcp
Start the server before using Continue. Or use cn --mcp http://localhost:3100/mcp.
Connect from Cursor
Option A – Command (recommended) – Cursor spawns the server; no manual start needed:
{
"mcpServers": {
"obsidian-cli-mcp": {
"command": "python",
"args": ["-m", "src.server"],
"cwd": "<project-root>",
"env": { "MCP_TRANSPORT": "stdio" }
}
}
}
Use your actual project path for cwd, or omit it if the project is your workspace root.
Option B – URL – Start the server manually first (python -m src.server), then add:
{
"mcpServers": {
"obsidian-cli-mcp": {
"url": "http://localhost:3100/mcp"
}
}
}
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。