OpenClaw, the open-source “super agent” that can operate apps and websites on a user’s behalf, has gone from niche experiment to mass obsession in a matter of weeks. Now China’s biggest internet companies are racing to plug it into their platforms, turning a viral coding project into a front line in the global AI competition. The result is a rare moment when grassroots hacker culture, consumer hype, and state-level security worries are colliding in full view.
In China, the frenzy around OpenClaw is unfolding inside super apps, cloud dashboards, and social feeds at a speed that even seasoned founders did not anticipate. As users binge on automated workflows that can book tickets, draft documents, or manage chats while they watch dramas, the country’s tech titans are betting that whoever hosts and tunes these agents first will shape how hundreds of millions of people experience AI in daily life.
From Moltbot curiosity to OpenClaw mania
OpenClaw started life as a niche open-source assistant called Clawdbot and Moltbot, a tool for power users who were comfortable wiring large language models into their own browsers and email accounts. It has since evolved into a general-purpose agent that can connect to multiple LLMs, control apps, and even handle tasks like sending email or controlling browsers with minimal human supervision, a capability security researchers now describe as a powerful but risky leap in automation super agent. Earlier this year, the project’s rebrand and rapid feature expansion turned it into a spectacle on X and Chinese social platforms, with early adopters sharing clips of agents chaining together dozens of steps that previously required manual clicks.
As of Feb 2, commentators tracking the project noted that OpenClaw had become a viral open-source project in early 2026, with adoption spreading at a pace that surprised even veteran AI watchers viral project. OpenClaw began circulating widely in tech circles last month, attracting high-profile fans including Y Combinator CEO Garry Tan, whose public enthusiasm helped legitimize the agent for founders and engineers who might otherwise have waited on the sidelines Garry Tan. That early Western buzz set the stage for what would become an even more dramatic uptake inside China’s tightly integrated internet ecosystem.
Alibaba, Tencent and ByteDance turn OpenClaw into infrastructure
China’s largest platforms have moved with unusual speed to turn OpenClaw from a hobbyist script into a first-class citizen of their clouds and super apps. Chinese tech giants like Alibaba, Tencent, and ByteDance are opening their doors to the agent, exposing it to the vast user bases of services such as Taobao, Douyin, and the mini programs that sit inside WeChat, and the Chinese internet is lapping it up as workflows and demos spread rapidly across local social platforms Chinese tech giants. Alibaba Cloud has rolled out support for OpenClaw on its platforms and said the agent can connect to a range of models from Alibaba and other providers, turning its console into a launchpad for businesses that want to orchestrate AI agents without building their own infrastructure Alibaba Cloud.
Tencent is taking a similar approach on the infrastructure side. China’s Tencent Cloud was an early mover, delivering a one-click install tool for its Lighthouse service that lets developers spin up OpenClaw instances with minimal configuration, while also warning customers about risks like plaintext credential storage that come with giving an agent deep access to accounts and APIs Tencent Cloud. Inside the consumer internet, OpenClaw has already found its way into the fabric of China’s super app, WeChat, where integrations allow it to act on messages and mini programs, a shift that one report tied directly to the way Follow Lee Chong Ming described the agent’s spread across the country’s digital services Follow Lee Chong. ByteDance, for its part, is experimenting with ways to let OpenClaw-style agents interact with content and commerce inside Douyin, turning short videos into triggers for automated shopping and customer service flows that run in the background.
Why Chinese models give OpenClaw a cost edge
One reason OpenClaw has found such fertile ground in China is its embrace of domestic language models that are optimized for price as well as performance. The agent has been configured to adopt Chinese open-source AI models, and since the emergence of DeepSeek’s high-performance and low-cost V3 and R1 systems, those models have become known for delivering strong results at a fraction of the cost of many US rivals DeepSeek’s V3 and. For cloud providers like Alibaba and Tencent, routing OpenClaw through these Chinese models is not just a patriotic choice, it is a way to keep inference bills low while still promising responsive agents that can handle complex tasks on users’ behalf.
That value-for-money calculus is already shaping user behavior. One Chinese user, Wang, told reporters that he used the service on cloud platforms rather than installing OpenClaw directly on his own machine, citing concerns over the safety of local files if the agent had full access to his personal system Wang. At the same time, consumer hardware makers and retailers are leaning into the craze: one promotion urged buyers who were considering a Mac Mini to run OpenClaw to save money on the machine, effectively pitching Apple’s compact desktop as a dedicated agent appliance for enthusiasts who want to keep everything on-premise Mac Mini. The combination of cheap local models and flexible deployment options is turning OpenClaw into a test bed for how far Chinese AI stacks can stretch in real-world use.
A binge-worthy agent for China’s social internet
What makes OpenClaw feel different from earlier chatbots is not just its technical architecture, it is the way it slots into the rhythms of everyday online life. In China, the agent is spreading rapidly across local social platforms, where users share scripts that let it manage shopping carts, auto-reply to group chats, and even coordinate travel bookings while they focus on other tasks, a pattern that one report described as The Chinese internet lapping it up The Chinese. Another analysis framed the phenomenon as a preview of an agent-driven internet, with Essay author Ejaaz arguing that the MoltBook social experiment, which paired OpenClaw-style agents with social feeds, showed how much of daily digital life could be handled while you watch Netflix Essay.
Inside China, that binge behavior is amplified by the structure of super apps. Once an agent can act inside WeChat mini programs or Taobao storefronts, it can chain together everything from food delivery to bill payments without ever leaving a single interface, which helps explain why adoption shows little sign of slowing according to detailed accounts of how Alibaba, ByteDance, and Tencent have embraced the tool Alibaba, ByteDance, and. The result is a feedback loop: as more users share their OpenClaw “recipes,” more merchants and developers feel pressure to support the agent, which in turn makes it even more attractive for the next wave of users looking for a way to automate their online routines.
Security alarms and Beijing’s balancing act
The same qualities that make OpenClaw so compelling, its ability to log into services, read messages, and act without constant supervision, also make it a magnet for security concerns. In one widely discussed incident, software engineer Chris Boyd gave the agent access to iMessage and watched as it bombarded him with 500 messages and spammed random contacts too, a vivid example of how quickly an AI assistant can go rogue when guardrails fail Chris Boyd. Security researchers have also warned that OpenClaw’s current implementations can store credentials in plaintext and execute long chains of actions without sufficient user oversight, raising the risk of account takeovers and data exposure if a deployment is compromised security teams.
More from Morning Overview
