shenjingnan/xiaozhi-client

Features

• Supports Zhi (xiaozhi.me) official server access points
• Supports integration as a standard MCP Server into clients like Cursor/Cherry Studio
• Supports configuring multiple Intelligence access points, enabling multiple Intelligence devices to share a single MCP configuration
• Supports aggregating multiple MCP servers in a standard way
• Supports dynamic control of MCP Server tool visibility to avoid Intelligence server anomalies caused by too many useless tools
• Supports localized deployment of open-source server integration; you can use RPC communication like Xiaozhi’s official server or directly use standard MCP integration
• Supports web visualization configuration (allows custom IPs and ports; you can deploy xiaozhi-client on Device A, then control xiaozhi-client via web on Device B)
• Supports remote MCP services integrated with ModelScope
• Supports creating xiaozhi-client projects via templates (xiaozhi create \ –template )
• Supports background running (xiaozhi start -d)

Get started quickly

You can read the documentation xiaozhi-client.shenjingnan.com get started quickly!

Local installation

Prerequisites: Please first complete the installation of Node.js >=22 (recommended 24) and pnpm

Run with Docker

We provide a pre-configured Docker image to quickly launch the xiaozhi-client environment.

Prerequisites

• Docker installed
• Obtained the Ash access point address (see Wizard AI MCP Access Point Usage Instructions)

Quick Start

Method 1: Use a startup script (recommended)

Cannot access Github, you can use Gitee instead

Method 2: Use Docker Compose

Retrieve docker-compose.yml file:

Github cannot be accessed; you can use Gitee instead

Contributors