© ROOT-NATION.com - Use of content is permitted with a backlink.
Xiaomi has announced the launch of a new experimental AI project called Xiaomi miclaw. The tool was unveiled today and aims to turn a smartphone into a much more autonomous AI assistant capable of performing tasks independently by interacting with various applications and system functions.
The company explains that Xiaomi miclaw was created as an early test product based on its own large-scale MiMo speech model. Unlike traditional digital assistants, which mostly answer questions or generate text, the new system should recognize the user’s intention and complete tasks independently using available tools on the smartphone.
After receiving permission from the user, the AI gains access to the device’s system functions, as well as to compatible third-party applications, allowing it to execute commands directly through them. In fact, this means that the assistant determines which tools to use and what actions to take to achieve the desired result. Honor smartphones with AI Agent are already using a similar approach, which can perform similar operations.
If a request requires opening a program, checking system data, or activating a certain function, AI independently determines the sequence of steps and executes them one by one. Xiaomi notes that the system can also work with less specific requests, trying to interpret them and turn them into concrete actions.
The platform is based on a mechanism Xiaomi calls the “inference and execution cycle”. First, AI analyzes the user’s request and then selects the appropriate tool and parameters for the action. Then, the system executes the command, checks the result, and, if necessary, continues the process until the task is completed. All these stages work asynchronously, so the assistant does not block other smartphone processes.
Xiaomi miclaw also has a memory system that should help AI adapt to repeated use. The assistant retains the key context of interactions and at the same time compresses old records to maintain an understanding of the user’s original intent even in long processes.
The system can also interact with the wider Xiaomi ecosystem. By integrating with the Mi Home platform, AI is able to receive information about the status of smart home devices and send commands to control them if the user provides appropriate access.
The company also opens the platform for developers. Support for the Model Context Protocol (MCP), an open standard for integrating AI tools, may allow Xiaomi miclaw to use existing AI solutions created for other platforms. Besides, Xiaomi is preparing an SDK that will allow third-party applications to inform the system about their capabilities so that AI can call them at the right time.
The company emphasizes that miclaw is still an experimental project. The issues of reliability, power consumption, and accuracy of complex tasks are still being improved, so some operations may be unstable or fail.
For this reason, the launch is in a very limited format. Xiaomi is opening miclaw only as a closed beta program, which is only available by invitation. The company also advises testers not to install the experimental version of the system on mainstream smartphones and to make sure to back up data before using it. The test currently supports a limited number of devices. The list includes Xiaomi 17, Xiaomi 17 Pro, Xiaomi 17 Pro Max, Xiaomi 17 Ultra, and Xiaomi 17 Ultra Leica Edition.
Xiaomi also notes that user interaction data with miclaw is not used to train AI models. According to the company, algorithms are trained based on open or licensed datasets, while personal interactions are used only to execute commands in real time. Confidential information is processed locally on the device using a technology Xiaomi describes as “edge cloud privacy computing”.
Read also:
- HyperOS 3.1: Xiaomi is preparing a revolution in Apple compatibility
- HyperOS is going on vacation: Xiaomi suspends updates until spring