FAQ
XiaoZhi AI primarily supports ESP32 series chips, including ESP32, ESP32-S3, and ESP32-C3. The community is also expanding support for Sifli, Allwinner, Quectel and other platforms. ESP32-S3 is recommended for the best experience.
Not at all! If you just want to build one, simply buy the hardware, use the web-based flash tool to install firmware, and configure the network — no coding required. For custom development, you'll need some C/C++ and embedded development experience.
XiaoZhi AI supports multiple large language models through the server, including Qwen, DeepSeek and more. You can switch between different models in the server configuration.
XiaoZhi AI supports offline voice wake-up with the default wake word 'XiaoZhi XiaoZhi'. The wake-up function runs locally on the device without internet. You can also customize the wake word by modifying the firmware.
MCP (Model Context Protocol) is the communication protocol used by XiaoZhi AI for device-server interaction. Through MCP, you can enable smart home control, knowledge search, and other extended capabilities.
Yes! The community provides server implementations in Python, Java, Go and more. Choose the version that suits you best for self-hosting, or use the official cloud service.
Contributions are welcome! You can submit Issues and Pull Requests on GitHub, participate in community discussions, and contribute code, documentation, hardware designs and more. The project uses the MIT open-source license.
Common causes include: USB driver not installed, serial port occupied, or wrong chip model selected. We recommend using the web-based flash tool, ensuring your browser supports WebSerial (Chrome recommended), and checking USB connection stability.