# MedXpert 图书馆连接器（通用版）

> MedXpert 图书馆连接器（通用版） is an open-source MCP server for knowledge and memory. Local MCP server that searches and answers questions over your own Markdown wiki. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://medxpert.cn
- **Repository:** https://github.com/zhaoxinghua09-cell/medxpert-library-connector
- **MCP Registry name:** io.github.zhaoxinghua09-cell/medxpert-library-connector
- **Categories:** knowledge and memory
- **Listing updated:** 2026-11-15

## About

This connector turns a folder of plain .md files into a searchable knowledge base that runs entirely on your computer. Retrieval and answers use a local Ollama instance (bge-m3 for embeddings and qwen2.5:7b for generation by default), so it needs no API keys, works offline and makes no external network calls. It is written with the Python standard library only.

It ships with a three-page sample wiki, and you point LIBRARY_WIKI_DIR at your own directory. Build the vector index with rag_engine.py, then start the stdio server. Answers from ask_knowledge include citations to the retrieved pages. The package is published as medxpert-library-connector on PyPI under Apache-2.0.

The publisher, MedXpert, is a Chinese medical-device regulatory consultancy, and this is its general-purpose edition; the project documentation is bilingual in English and Chinese.

## Features

- **Local semantic search:** Searches your wiki with embeddings from a local Ollama model.
- **RAG answers:** Combines retrieval and a local model with citations.
- **No external calls:** Runs offline; queries and documents stay on your machine.
- **Standard library only:** No third-party runtime dependencies.
- **Configurable models:** Environment variables set the wiki folder, Ollama URL and models.
- **Sample wiki:** Includes three example pages to test with.

## Good for

- Search personal Markdown notes by meaning
- Ask questions of an internal wiki with cited sources
- Read one wiki page in full
- Check the size and categories of a knowledge base
- Run retrieval in an air-gapped environment

## Install

### Claude Code (Terminal)

```sh
claude mcp add medxpert-library-connector -- uvx medxpert-library-connector
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "medxpert-library-connector": {
      "command": "uvx",
      "args": [
        "medxpert-library-connector"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "medxpert-library-connector": {
      "command": "uvx",
      "args": [
        "medxpert-library-connector"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "medxpert-library-connector": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "medxpert-library-connector"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "medxpert-library-connector": {
      "command": "uvx",
      "args": [
        "medxpert-library-connector"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "medxpert-library-connector": {
      "command": "uvx",
      "args": [
        "medxpert-library-connector"
      ]
    }
  }
}
```

## Tools

- `search_knowledge`: Semantic search over your wiki, returning snippets with similarity.
- `get_page`: Returns the full text of one wiki page.
- `list_sources`: Shows page count, categories and index status.
- `ask_knowledge`: Answers a question using retrieval plus a local model, with citations.

## FAQ

### Does it send my documents anywhere?

No. The README says everything runs locally and no data is uploaded.

### What do I need installed?

Python and Ollama with an embedding model; a generation model is optional for answers.

### How do I use my own notes?

Set LIBRARY_WIKI_DIR to a folder of .md files and build the index.

### What licence is the code under?

Apache-2.0.

### Is MedXpert 图书馆连接器（通用版） free?

Yes. MedXpert 图书馆连接器（通用版） is open source, so you can use and self-host it for free.

### How do I add MedXpert 图书馆连接器（通用版） to Claude Code?

Run this command in your terminal: claude mcp add medxpert-library-connector -- uvx medxpert-library-connector

### What are the best alternatives to MedXpert 图书馆连接器（通用版）?

The closest alternatives to MedXpert 图书馆连接器（通用版） are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.

## Alternatives

- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Nautilus Compass Memory](https://ai-powered.com/mcp/nautilus-compass-memory): Nautilus Compass Memory is an open-source MCP server for knowledge and memory.
- [note1](https://ai-powered.com/mcp/note1): note1 is an MCP server for productivity.
- [Context Nest](https://ai-powered.com/mcp/context-nest): Context Nest is a freemium MCP server for knowledge and memory.
- [Tulsa Ecosystem Map](https://ai-powered.com/mcp/tulsa-ecosystem-map): Tulsa Ecosystem Map is a free MCP server for knowledge and memory.
- [Agent Yurt](https://ai-powered.com/mcp/agent-yurt): Agent Yurt is a free MCP server for knowledge and memory.

---
Source: https://ai-powered.com/mcp/medxpert
