Build Your OwnXiaoZhi AI Device
An open-source AI chatbot built on the MCP protocol. Powered by LLMs like Qwen and DeepSeek, enabling ultra-low latency natural voice interaction and smart device control on ESP32.
Choose Your Path
Whether you want a quick hands-on experience or deep custom development, we have comprehensive guides for you.
Quick Build
For beginners & enthusiasts
Assemble with a breadboard and components, flash official firmware, connect to cloud services — just 4 steps to your own AI voice companion.
Deep Development
For makers & developers
Custom hardware design, compile firmware from source, extend MCP protocol capabilities — build a version that's entirely yours.
System Architecture
From hardware to cloud LLMs, a complete and ultra-fast technology pipeline.
Hardware
MCU + Microphone + Speaker + Display
Firmware
Wake Word + Audio Codec + Communication
Protocol
WebSocket / MQTT / MCP
Server
LLM + TTS + AI Agent + Knowledge Base
Core Capabilities
XiaoZhi AI is more than a voice assistant — it's a complete AI + IoT development platform.
Natural Voice Interaction
Streaming ASR + LLM + TTS architecture with offline wake word detection and voiceprint recognition for ultra-low latency conversation.
MCP Protocol Control
On-device MCP controls speakers, LEDs, servos and more. Cloud MCP extends smart home and knowledge search capabilities.
Multi-Chip Platform
Full support for ESP32-S3, ESP32-C3, ESP32-P4 and other Espressif chips, plus Sifli, Allwinner, Quectel platforms.
Rich Display
Supports OLED / LCD displays with dynamic Emoji expressions, conversation content, custom fonts and background UI.
Fully Open Source
MIT licensed, 24.9K Stars, 110 contributors, with an extremely active community ecosystem and vast third-party firmware.
Multilingual Support
Native support for Chinese, English, Japanese and more — easily adaptable for global developers and localization needs.