In This Story
In the past year, artificial intelligence leaders have talked about the advent of agents that can do work autonomously. Now, China says it has developed the world’s first.
Last week, Chinese researchers launched an early preview of Manus AI, which they said is “the first general AI agent.”
“This isn’t just another chatbot or workflow,” Yichao “Peak” Ji, chief scientist for Manus AI, said in an introductory video. “It’s a truly autonomous agent that bridges the gap between conception and execution.”
While other AI agents are useful for idea generation, Ji said Manus AI “delivers results” without much human prompting. The agent’s name comes from the Latin motto “Mens et Manus,” which translates to “Mind and Hand.”
“We see it as the next paradigm of human-machine collaboration, and potentially a glimpse into AGI,” Ji said, referring to artificial general intelligence — the point when AI will be believed to have reached and surpassed human-level intelligence.
The agent is currently invitation-only, and access codes were being resold for thousands of dollars on China’s reseller app, Xianyu (BABA-0.44%), TechCrunch (VZ-0.27%) reported. Manus AI did not immediately respond to a request for comment from Quartz.
Here’s what we know so far about Manus.
What can Manus do?
According to the Manus AI website, the agent can perform real-world tasks such as providing custom travel plans, researching real estate properties for affordability, and performing correlation analyses between stocks.
Ji demonstrated Manus screening 15 resumes, then providing its ranking suggestions and evaluation of each candidate. He then instructed the agent to put the information into a spreadsheet.
“Manus has its own knowledge and memory, so I can teach Manus that the next time it handles a similar task, it will deliver a spreadsheet right away,” Ji said
Because Manus works asynchronously in the cloud, Ji said, users can close their laptop while it completes tasks. The agent can also receive new instructions while it is working.
On the GAIA Benchmark, which evaluates general AI assistants, Ji said Manus has achieved state-of-the-art performance alongside OpenAI’s Deep Research agent.
The agent is already performing tasks on freelance work platforms such as Upwork (UPWK+1.62%) and Fiverr (FVRR+1.87%), Ji said.
How was Manus built?
Ji said the agent’s capabilities “wouldn’t be possible without” the open-source community, which means its code, datasets, and parameters are available for anyone to access and build upon.
The agent “operates as a multi-agent system,” and is powered by different AI models, Ji said, adding that the team plans to open-source some of the models later this year.
According to Hugging Face, Manus was developed by a Chinese AI startup called Monica.im, which is developing next-generation autonomous agents. However, other reports say Manus was built by a Chinese firm called Butterfly Effect.
What are people saying about Manus?
Dean Ball, an AI research fellow at George Mason University, said in a post on X that it was “wrong” to compare Manus to China’s breakthrough AI moment with DeepSeek earlier this year.
“Deepseek was about replication of capabilities already publicly achieved by American firms,” Ball said. “Manus is actually advancing the frontier.”
Victor Mustar, head of product at Hugging Face, called Manus “the most impressive AI tool I’ve ever tried,” in a post on X. Mustar said Manus’s “agentic capabilities are mind-blowing, redefining what’s possible.”
However, other AI researchers were not so impressed.
Alexander Doria, co-founder of French AI lab PlelAs, said in a post on X that despite liking the agent’s user-interface, “it’s fundamentally a workflow” and “not an actual agent (at least nothing really beyond the built-in agentic capacities of Claude).”
Professor and researcher Derya Unutmaz said in a post on X that he ran OpenAI’s Deep Research alongside Manus. While Deep Research completed Unutmaz’s task in under 15 minutes, Manus failed after 50 minutes, and didn’t finish all the necessary steps.
