Xiaomi officially announced the limited closed beta testing of Xiaomi miclaw today in China, marking a major milestone in AI agent technology. Built on the Xiaomi MiMo large model, this tool allows users to control their Xiaomi smartphones and smart home devices with single-sentence commands.
The agent understands complex intent, remembers user habits, and can even suggest ways to save roughly $55 annually by analyzing subscription bills. This invitation-only test is currently available for the latest Xiaomi 17 Ultra and Pro models.
What is Xiaomi miclaw?
Xiaomi miclaw is an advanced AI interaction product designed to transform a standard smartphone into a proactive digital agent. Unlike traditional voice assistants that follow simple triggers, this system is built on the Xiaomi MiMo large model to understand human intention and execute multi-step workflows. It functions as a core system application that can autonomously call upon over 50 different system tools and ecosystem services to complete complex tasks efficiently.
The software is designed to grow more effective the more a person interacts with it, thanks to its sophisticated memory system. By analyzing authorized data, such as calendar events or SMS notifications, it can make intelligent decisions like muting a phone during a meeting or pausing a robot vacuum. It represents the next step in the Xiaomi HyperConnect ecosystem, bridging the gap between mobile software and IoT hardware through natural language.
How Xiaomi miclaw works
The technical architecture of Xiaomi miclaw relies on an advanced inference-execution loop that processes user input through a continuous feedback cycle. The system analyzes the user’s voice or text command, selects the appropriate tool, and executes the action while monitoring the results in real-time.
This process occurs asynchronously, ensuring that the system remains responsive even while handling complex background calculations or communicating with various IoT devices via Xiaomi HyperConnect protocols.
-
Autonomous Task Execution: Xiaomi miclaw can independently select system-level tools and third-party applications to complete vague commands. After receiving authorization, the AI interprets user needs and executes multi-step workflows without manual intervention.
-
Three-Tiered Memory Management: The core engine uses an advanced memory system that retains key decision points and compresses interaction history. This allows the AI to maintain context over 20 consecutive steps, ensuring it understands the initial goal throughout long tasks.
-
Comprehensive IoT Control: By integrating the full Mi Home protocol, the agent translates machine-readable specifications into natural language descriptions. This allows it to manage air conditioners, lighting, and security systems within the smart home network.
-
Supported Devices: This initial beta test is exclusive to the Xiaomi 17 Ultra Leica Edition, Xiaomi 17 Ultra, Xiaomi 17 Pro Max, Xiaomi 17 Pro, and the standard Xiaomi 17.
-
Advanced Developer Integration: The software supports the Model Context Protocol (MCP) and a dedicated Open SDK for third-party developers. Apps can proactively declare their capabilities, allowing the AI to discover and use new functions dynamically.
-
Strict Privacy Standards: Xiaomi confirms that personal data from these interactions will not be used for model training purposes. Most processing occurs locally via edge-cloud privacy computing to keep sensitive information secure on the device.
This beta phase will continue to optimize power consumption and success rates before a wider public release.
Via: Xiaomi

Emir Bardakçı

