On June 25, 2026, Xiaopai Technology released the first home smart terminal ClawStation in Shenzhen, putting a 7B large model and four-channel visual perception directly into the home smart home center.
Different from the mainstream solutions in the industry that rely on infrared and millimeter wave radars for passive sensing, this product directly uses local AI vision, allowing smart homes to truly “understand” what is happening at home for the first time.
The pain point of smart homes is never control, but perception.
Why do we still need to manually shout commands to trigger actions after using smart homes for so many years?
What’s the real problem?
The industry dilemma of passive sensing is never lacking sensors
You must have encountered such a scene: when you wake up at night, the human body sensor detects someone, and the lights in the living room are turned on so brightly that it is dazzling; when a guest comes, the sensor cannot tell whether it is a family member or a visitor, and the air-conditioning temperature remains at the 26 degrees you are used to; when a paper towel falls on the ground, the sensor cannot see it at all, and the sweeping robot can only walk around.
This is the Achilles heel of traditional smart homes: all sensors can only do “presence detection” and cannot do “semantic understanding.”
Infrared and millimeter wave radars can’t lie, but they can only tell you “there is something here”, but they cannot tell you “who this thing is and what it is doing”.
The industry has actually been aware of this problem for a long time, but no one has dared to use cameras as main sensors on a large scale. The core issues are two: privacy and computing power.
Users are worried about transmitting home images to the cloud for analysis; it relies entirely on end-side processing, which requires insufficient computing power and high costs.
As the open source agent framework expands application scenarios, computing power is shifting from the cloud to the end-side. The two major trends of streamlining multi-modal large models and improving chip performance and reducing power consumption are superimposed. The implementation of home end-side visual AI is finally feasible.
This time ClawStation Pro directly provides a landing plan: locally deploy a multi-modal large model with 7B parameters, and use the “main control module + AI computing power card” architecture to achieve local inference. Official data shows that the inference accuracy is increased by 16%, the speed is increased by 11%, and the power consumption is reduced by 37%.
All video analysis is completed locally, and the data does not need to leave the house, directly solving privacy concerns.
The core of visual intelligence is to understand “family needs”
ClawStation Pro can connect to four cameras, and all analysis is completed on the end side.
This means that it can not only “see” but also “understand”: it can distinguish whether there are elders or children in the house and automatically adjust the air conditioning temperature; when it sees garbage on the ground, it directly sends instructions to the sweeping robot to clean it; it recognizes that the elderly are looking at their mobile phones for three consecutive hours and will send reminders to their children.
Unlike the care products on the market that only target a single group of people, Xiaopai positions ClawStation as a full-scenario intelligent body for the entire family.
This positioning just hits a gap in the industry – there are now baby care cameras and pet smart devices, but there are few integrated solutions for the entire family, especially those focusing on the care of the elderly living alone.
- For the elderly living alone: You can set reminders for getting up, going out, eating, and abnormal activity levels, and generate daily activity summaries
- For children: You can monitor the time you get home from school, homework time, and TV watching time, and automatically remind you when the limit is exceeded
- For pets: You can pay attention to eating patterns, abnormal activities, and prevent pets from climbing on the dining table
The dedicated care terminal “KinSight” is currently being piloted in some elderly care institutions. It can set care rules according to user guidelines and generate daily care reports.
Some users reported that the device once pushed a reminder saying, “The old man is doing well today, but he has been looking at his phone for three hours.” This kind of detail is simply not possible with traditional sensors.
Xiaopai also admitted that home intelligence is still in its infancy. The recognition cannot be 100% accurate, and the decision-making cannot be completely error-free. Users need to “teach it slowly and cultivate it slowly.”
This pragmatic attitude is more reliable than the slogan of “subverting everything”.
The ambition of open interfaces is not only to make products but also to make ecology
This press conference simultaneously announced another major move: Xiaopai became the first manufacturer in the industry to fully open the camera device interface, providing standard skills such as device search, security certification, audio and video acquisition, and device settings. All capabilities have been put on the ClawHub skills store.
Users can use any intelligent agent such as OpenClaw and HermesClaw to access their own cameras and DIY their own personalized application scenarios, which is equivalent to returning “the power to define scenarios to the user.”
This action also corresponds to Xiaopai’s strategic upgrade: from the original audio and video smart terminal supplier to a smart home solution provider integrating cloud, edge and terminal.
Open interfaces are not for charity, but to occupy ecological standards in advance when the home intelligent industry is just starting out.
The PaiClaw agent framework is developed based on OpenClaw and is fully compatible with the latter’s instructions, skills and ecology. It is only adapted to family scenes and visual perception.
For developers and geek users, this means that existing ecological resources can be directly reused without reinventing the wheel.
This open idea just hits the current pain point of the smart home industry: it is difficult for devices of various brands to interconnect and interoperate, and it is even more difficult for users to combine their own scenarios.
Xiaopai’s opening of the camera interface is equivalent to opening a new door for all developers who want to create smart home scenarios.
The essence of family AI is protection, not replacement
A point of view mentioned at the press conference was recognized by many people in the industry: Alan Turing asked “Can machines think?”. Now the new question that home AI needs to answer is “Can machines feel love?”.
We create it not to replace human intelligence, but to protect human emotions. This is the meaning of AI’s existence.
This sentence actually reveals the essence of family AI.
Over the years, the smart home industry has been comparing parameters and the number of connected devices, but few people have truly returned to the “home” scene itself – what users want is not a bunch of devices that can connect to the Internet, but an assistant that can truly share the pressure of family care.
Especially now that the population is aging, children cannot stay with the elderly all the time, young parents are busy with work and cannot keep an eye on their children around the clock, and the number of people raising pets is growing. Full-scenario home care itself is a real need.
Xiaopai chose this direction to hit the turning point of social needs.
It is understood that the ClawStation series is not currently open to ordinary consumers for purchase, and the actual recognition accuracy and intelligent capabilities will not be verified until it is launched.
But it is undeniable that this solution of using local visual AI for active sensing has indeed pointed out a new direction worth exploring for the smart home industry.
When AI really comes into the home, it does not need to replace human companionship. It only needs to catch those details that we do not see or care about, and help us implement our unspoken concerns into our daily lives. This may be the most touching appearance of home intelligence.
