Alibaba has a new bet on the future of work, and it runs like a mule. MuleRun, the agentic AI platform developed by Alibaba Group Holding, has quietly grown into a multi-country productivity service that its creators describe as an “always-on AI workforce,” a system that continues executing tasks even when the user has stepped away from their screen. According to SCMP reporting published on May 28, the platform now serves businesses and users in 43 countries, including China, Brazil, and Mexico, and among its paid users, more than a third spend over US$200 per month.
The name is deliberate. MuleRun’s team chose it to reflect the idea of AI agents as mules, reliable, tireless workers that handle the “repetitive and trivial” tasks that consume human time and attention. The platform was initially launched last September as an AI agent marketplace before transitioning into a broader AI productivity platform. It was formally showcased at the Alibaba Cloud summit on May 20.
The Team Behind MuleRun
MuleRun was developed by a team led by Chen Yusen, vice-president at Alibaba Cloud. Chen is a graduate of Zhejiang University, the same institution that produced DeepSeek founder Liang Wenfeng, a detail that underscores the concentration of AI talent flowing from a handful of Chinese universities into the country’s leading AI ventures. Chen joined Alibaba in 2023, initially as a researcher in the CEO’s office before moving into a role overseeing global expansion, including a stint in Mexico City that gave him direct exposure to the Latin American market that MuleRun now serves.
The platform’s agent roster covers a range of commercial use cases: video generation, stock analysis, social media management, research, report writing, and coding. Users can schedule workflows to run overnight, meaning that a task assigned at the end of a working day can be completed and waiting in the morning, the “always-on” quality that distinguishes agentic AI from conventional chatbot interactions.
MuleRun vs. OpenClaw: The Safety Argument
The most pointed element of the SCMP coverage is MuleRun’s positioning against OpenClaw, Tencent’s competing AI agent framework. OpenClaw requires users to download software and grants the application broad system permissions, a security exposure that MuleRun’s team argues creates unacceptable risks for enterprise users. MuleRun, by contrast, runs entirely through natural-language prompts in a browser-based environment. No software download is required, and no elevated system permissions are granted.
This distinction matters because enterprise AI adoption is frequently blocked not by capability concerns but by security and IT governance concerns. A platform that can be accessed without touching a company’s endpoint security posture is easier to deploy at scale, particularly in regulated industries. Among MuleRun’s paid users, the average of 13 end-to-end tasks completed weekly suggests that users who adopt the platform integrate it deeply into their workflows rather than using it occasionally.
As EastFrontier has reported, Alibaba Cloud AI revenue hit 9 billion yuan in Q1, but the monetization gap remains wide. MuleRun represents one of Alibaba’s clearest attempts to close that gap by converting AI capability into recurring subscription revenue from enterprise and professional users. The platform’s expansion to 43 countries also reflects Alibaba’s broader push to build an international AI business at a time when Chinese tech giants are dangling record pay to court Singapore AI graduates and expanding aggressively across Southeast Asia and Latin America.
The Agentic AI Race in China
MuleRun’s launch comes as the agentic AI category, platforms that execute multi-step tasks autonomously rather than simply answering questions, is emerging as the next major battleground in Chinese AI. Alibaba’s Qwen3.7-Max ranked 4th on the global Code Arena leaderboard in part because of its tool-calling capabilities, which allow it to orchestrate complex workflows. MuleRun is the consumer and enterprise face of that underlying capability — the product layer that turns model performance into commercial value.
The platform’s 43-country footprint also positions Alibaba ahead of most Chinese AI companies in the race to build international distribution for agentic products. As the AI agent market matures, distribution and trust, not just model capability, will determine which platforms become the default infrastructure for AI-assisted work. MuleRun’s browser-based, no-download architecture is a deliberate answer to that trust problem: it asks less of the user and, in doing so, removes the primary friction point that has slowed enterprise AI adoption in security-conscious organizations.