What does a home robot connected to an AI agent still lack to become a real "family member"?

How deep an emotional connection can be established between humans and silicon-based life? When robots were "sold out" again in the first quarter of 2026, this question was not resolved as the industry heated up; instead, it became more pressing.

Today's embodied intelligence industry has entered a more specific and brutal stage: Many products that performed well in the laboratory cannot understand users' personalized needs once they enter the real living environment. They may be able to execute single commands but cannot complete long-term tasks.

This actually determines the future direction of the entire industry: Will future household robots merely be mass-produced AI toys, or will they have the opportunity to become true friends of humans, like in the story of "Robot Dreams"?

In the process of integrating AI Agents, LeXiang Technology, the global pioneer in consumer-grade embodied intelligence, saw new possibilities. After integrating Tencent Cloud's ClawPro, the average daily interaction time of LeXiang's internal test users increased from 30 minutes to 2 hours. Users would continuously add new skills to test what else the robot could do; they would let the robot try to control other household appliances through IoT; they would tease or even provoke it. Even a stock trader would associate it with his stock performance. When the stocks rose, the robot would dance for him.

"Why did people stop playing with a product after just two days in the past? Because it was 'imprisoned,'" said Li Yuanqing, the co-CTO of LeXiang Technology, his eyes lighting up when talking about robots. "Once users saw all the possible ways to play, without new surprises and differences, they naturally couldn't have new emotional experiences." As one of the first robot companies to integrate ClawPro, LeXiang aims to bring different surprises to users.

Bringing infinite possibilities not written in the manual is the biggest variable that ClawPro has brought to the embodied intelligence industry.

01 From a Big Toy to a Real "Family Member"

Compared with the B2B industrial scenarios where robots are highly involved, the household scenario is highly dynamic, complex, and has sufficient room for growth. For a young founding team, it is the most promising ground in this era.

This is the reason why Li Yuanqing joined LeXiang without hesitation. Before that, he was the main person in charge of Huawei Cloud's embodied intelligence. This 90s genius split the attributes of household robots into three levels. The first level is "pet," which provides a certain amount of emotional value like a pet; the second level is "partner," offering half emotional value and half functional value; the third level is "assistant," capable of completing complex tasks that many traditional tools cannot.

Li Yuanqing, Co-CTO of LeXiang Technology

From any single aspect, some peers have already entered and made breakthroughs. However, what LeXiang wants to do is to explore by combining these three paths, making the robot a real "family member." This means that they are truly intelligent, capable of doing complex tasks, and able to understand the deeper intentions and needs of other family members, rather than just being a big toy for occasional chats. True intelligence is related to "whether we can combine the world knowledge, task reasoning, and command compliance that AI is truly good at in this era with strategies of high reliability, low cost, and high interpretability."

For example, a simple command from the user like "Check if my bedroom door is closed" may contain a wealth of information behind it. The robot needs to understand "Is it cloudy or rainy outside? Does it need ventilation? Is the cat in the room when closing the door? Is the air conditioner or the baseboard heater on or off? After the operation, does it need to call the user to confirm?" and so on.

This is no longer just simple voice recognition but a series of complex intention understanding and operations. First, it is a test of the robot's "perception" ability. At this time, the AI Agent capabilities of ClawPro can quickly call the context, enabling the robot to fully understand the user's needs. When the robot actually goes to "close the door," the series of AI Agent strategies set by ClawPro help build the robot's "reasoning" ability, break down task steps, mobilize different skill modules (skills), form IoT linkage, and finally execute specific actions (action).

That is to say, after integrating the AI Agent of ClawPro into LeXiang's robots, the path from "interacting with programs" to "interacting with people" has been truly opened up.

When robots truly participate in family life, unexpected needs begin to emerge, and even create different ways of combining skills. For example, during the internal test, elderly users would keep talking to LeXiang's robots, sharing their life experiences. The robots could be adjusted to be a "straight man" in a cross-talk, and even summarize the stories shared by the elderly for their children.

For robots, each execution is not only the completion of a task but also an opportunity for continuous optimization. Li Yuanqing summarized this by saying, "AI agents have completed the journey from 0 to 1, but from 60 to 100, everyone needs to create and polish together." Just as AGI evolves continuously to reach perfection, fully intelligent embodied intelligence does not exist at this stage. Only with the increase in interaction time and the number of executed tasks can robots become "smarter" family members.

Currently, LeXiang's three robots (M1, W1, N1) are still in the internal test stage and have not fully withstood market tests. However, Li Yuanqing has already seen the infinite possibilities of robots entering households.

02 Why Tencent ClawPro?

For consumer-grade embodied intelligence products, in addition to breakthroughs in technology and key hardware, they also need to face more real-world problems to fully enter households.

The most direct ones are safety and stability. Household robots are directly in the real family environment, facing the elderly, children, and pets. Every move directly affects "people," so the requirements for safety are higher.

However, during the training process of "ClawPro," unsafe factors are everywhere. For example, some viruses may be implanted in the prompt, directly affecting the robot's physical operations. Safety is also one of the core reasons why LeXiang chose Tencent ClawPro. This star embodied intelligence company, which has been established for less than two years, has a core R & D team from top universities such as Tsinghua University, Zhejiang University, and Carnegie Mellon University. They could have developed their own system to integrate with open claw. However, building the central scheduling resources, servers, and safety control systems required for AI Agents from scratch is still too heavy a burden for a startup team. Therefore, LeXiang decided to rely on Tencent Cloud and focus on expanding the robots' ability boundaries.

"Embodied intelligence is a long-term and promising track. We believe in the saying that one can go fast alone, but a group can go far together." Li Yuanqing is very certain that this is the right choice. LeXiang's trust in Tencent was established during the cloud usage stage, and the feedback from user interactions after integrating ClawPro has further strengthened this trust.

"We call it 'visible, auditable, scannable, and safe.' These are the core problems that ClawPro needs to solve," explained Zeng Fangang, the general manager of Tencent Cloud's Shanghai & Zhejiang Region, to 36Kr. For example, when robots need to call different skills to perform tasks, Tencent Cloud provides a highly secure skills market, allowing robot companies to use them without worry. These capabilities do not directly appear on the front-end interface, but they determine whether the robot is truly "reliable."

Zeng Fangang, General Manager of Tencent Cloud's Shanghai & Zhejiang Region

Moreover, compared with other hardware products, robots naturally have the advantage of integrating AI Agents. Robots are equipped with a large number of sensors, which can continuously collect information generated by the interaction between humans and the environment in the real world, and are more likely to expose new needs. As robots become more and more familiar with people's living habits and behavior rhythms, the interaction in real scenarios will, in turn, promote the continuous iteration of AI Agent capabilities, further polishing an ecosystem with low risk and high stability.

After the establishment of a stable underlying ecosystem, the LeXiang team can build up the technical capabilities for agile development in a lighter way.

In the past, the team often adopted the traditional hardware R & D model, where product managers first defined requirements and created PPTs, and then entered a long development cycle. After integrating Tencent ClawPro, the team no longer needs to repeat the underlying construction and can focus on responding to users' needs, testing out the R & D points of embodied intelligence through real feedback. At this time, LeXiang Technology has also shifted from the waterfall-style R & D to rapid demo-driven development. Li Yuanqing described this process as "building with Lego." First, quickly assemble a prototype with a bunch of ready-made building blocks. After verification, continue to polish it to truly develop the ability for agile development.

LeXiang Technology maintains an almost "technological romanticism" attitude towards the future opportunities in the embodied intelligence track. In Li Yuanqing's view, "Most of the time, this track is immature, but this immaturity is also an opportunity to widen the gap."

Today, the embodied intelligence industry, whether in terms of product form, interaction mode, or business model, is still in a long evolutionary process. At the same time, more and more enterprises that face immaturity directly and are brave enough to seize opportunities are emerging. Tencent Cloud's ClawPro can be a trustworthy and safe friend, accompanying enterprises to explore new opportunities in the intelligent world.