Posts tagged: opening

The viral AI agent OpenClaw — formerly known as Clawdbot, then Moltbot — has found an audience in China.

Since last week, Chinese tech companies including Tencent, Alibaba, and Volcano Engine, a cloud service platform under ByteDance, have begun integrating OpenClaw into their platforms, making it easier for Chinese users to run the agent. That includes connecting the agent to workplace tools such as Alibaba’s collaboration platform, DingTalk, and Tencent Holdings’ WeCom, the work version of China’s super app, WeChat.

OpenClaw began circulating widely in tech circles last month, attracting high-profile fans including Y Combinator CEO Garry Tan and multiple partners at Andreessen Horowitz.

The agent has also taken off among Chinese users, with demos, tutorials, and use cases spreading rapidly across local social platforms.

OpenClaw is designed to run around the clock and plug into a wide range of consumer apps, allowing users to automate tasks such as managing schedules, overseeing vibe-coding sessions, or even building AI employees.

In a post on Tencent Cloud’s developer platform, the company said last Thursday that its servers have rolled out a preconfigured OpenClaw application template, enabling users to deploy the AI assistant in the cloud with minimal setup.

Alibaba Cloud has also rolled out support for OpenClaw on its platforms and said the agent can connect to a range of models from Alibaba’s Qwen series.

Volcano Engine, ByteDance’s cloud services arm, outlined how developers can deploy Moltbot in its environment in an article published on Monday, while also flagging key safety considerations.

“Because the tool has extensive data, account, and network access permissions, please deploy it in a dedicated environment, avoid handling sensitive information, and be sure to review permissions regularly and set access restrictions for ECS and API keys,” the article said, referring to cloud servers and access credentials.

For OpenClaw to run as a digital assistant across apps, it requires access to users’ files, login details, browser activity, and other data.

Cybersecurity specialists told Business Insider in a report published on Wednesday that agents like OpenClaw can be vulnerable to “prompt injections,” a tactic that uses hidden instructions to trick AI into performing actions such as leaking data or publishing content on users’ behalf.

Despite mounting privacy and security concerns, enthusiasm for the agent among Chinese users shows little sign of slowing.

OpenClaw’s popularity on Chinese social media

Posts and demos featuring OpenClaw have surged on the Chinese social media platform RedNote.

One RedNote user who goes by “Brother C” posted a video tutorial last Tuesday, walking viewers through how to use OpenClaw. “See how the 24/7 proactive AI assistant is revolutionizing workflows,” he wrote. The post drew more than 4,000 likes and was saved over 6,000 times.

Another user posting under the nickname “Teacher Du” shared his own explainer on Monday, describing how OpenClaw could be deployed in everyday workflows. His post was saved more than 2,000 times and received over 1,000 likes.

“My experience was truly mind-blowing,” he wrote, adding that the agent could handle “all sorts of tasks” and that the “concept of a true AI employee is getting closer.”

Like their counterparts in the US, Chinese users are buying Mac Minis to run the agent. A RedNote user named Wu Bin said he had ordered a secondhand Mac Mini to serve as his “super assistant.”

“It’s incredibly convenient, I can control it remotely to organize files and handle all sorts of tasks,” he wrote.

Not everyone is convinced. A user who goes by “Programmer Yago” warned in a RedNote post on Sunday that using the agent could leave users’ data “running naked all over the internet.”

OpenClaw did not respond to a request for comment from Business Insider.