Librarian MCP Server
A local MCP server that gives AI agents structured access to a personal Obsidian knowledge vault, with semantic search, organization through Maps of Content, and git-backed history.
README
Librarian MCP Server
A local MCP server that gives AI agents structured access to a personal Obsidian knowledge vault. Notes remain portable Markdown, while Librarian enforces organization, provenance, deduplication, and git-backed history.
What it does
- Stores OKF-compatible Markdown notes in a flat Obsidian vault.
- Organizes concepts through Maps of Content (MOCs).
- Combines local semantic search with lexical matching and exact filters.
- Rejects duplicate concepts, invalid notes, and edits to immutable source notes.
- Commits every mutation to the vault's git history.
- Reports broken links, orphaned notes, malformed frontmatter, and WikiLinks.
All search and storage are local. The first semantic search downloads the BAAI/bge-small-en-v1.5 model (about 130 MB); later searches run offline.
Install
Requires Python 3.11+.
pip install .
Or with uv:
uv sync
Run
librarian --vault /path/to/vault
You can also set the vault path with LIBRARIAN_VAULT:
LIBRARIAN_VAULT=/path/to/vault librarian
On first run, Librarian creates the vault, initializes git, and adds a root MOC. Configure your MCP client to launch the server over stdio, for example:
{
"mcpServers": {
"librarian": {
"command": "librarian",
"args": ["--vault", "/path/to/vault"]
}
}
}
Tools
| Tool | Purpose |
|---|---|
get_vault_seed |
Return a bounded overview of MOCs and recent notes |
query_knowledge |
Search notes semantically and lexically |
read_concept |
Read a complete note |
create_concept |
Create a validated note and wire it into an MOC |
enrich_concept |
Append content to a named section |
lint_graph |
Report graph and formatting problems |
See spec.md for the full contracts and design decisions.
Development
uv sync --extra dev
uv run --extra dev pytest
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。