Xiaomi’s New Miclaw AI Agent Automates Complex Tasks Across Devices

The Chinese technology company Xiaomi has begun its experimental testing phase of Miclaw, which operates as an artificial intelligence system developed for their research into agent-based AI technology dedicated to smartphones and their connected devices. The tool has entered a limited closed beta phase, where selected users are evaluating its ability to complete multi-step digital tasks autonomously.

Miclaw enables direct device control through its system. The AI system can interact with apps, system tools, and connected services after receiving user authorization, allowing it to perform tasks rather than simply responding to queries.

The technology is powered by Xiaomi’s proprietary MiMo large language model, which allows the system to understand natural language instructions and translate them into operational steps. The AI system initiates its process by first examining the user command to find out which resources will help it achieve the complete task.

Miclaw is also designed to handle workflows involving multiple steps. The system enables users to execute a single command that functions as an all-in-one solution for sending SMS messages, creating calendar appointments, searching the internet, and determining travel times. The system can help users buy products online, while it will also adjust device configurations automatically based on the current situation of the user.

Another notable feature is its integration with Xiaomi’s wider ecosystem. The AI system establishes links to smart home devices and multiple networked items, which enable it to manage home electronics and perform functions based on how users interact with their devices. The assistant can make a smartphone silent during meetings, and it has the ability to stop a robot vacuum cleaner from operating when necessary.

The company currently conducts the closed beta programme on specific devices that belong to the Xiaomi 17 series that comprises the Xiaomi 17 Pro and Xiaomi 17 Ultra. The company says the testing phase aims to evaluate how well AI models can manage system-level operations across its “human-car-home” ecosystem.