Who Built Manus AI? The Startup Story Behind the Agent

In March 2025, a short demo video went viral. It showed an AI that did not just chat. Instead, it opened a browser, did the research, and handed back finished work. That was Manus AI. Within days, people were reselling invite codes for thousands of dollars.

So, who built it? The short answer is a small team led by Xiao Hong and Ji Yichao. However, the full story is far more compelling. It runs from WeChat tools to a browser plug-in, then to a viral launch. After that came a move to Singapore, a $2 billion deal with Meta, and a dramatic reversal.

In this listicle, you will meet the people behind Manus. You will also see how they built it and what founders can learn from their path.

The Founders: Who Built Manus AI?

To start, let’s meet the core team. Manus AI came from three people with very different strengths.

Xiao Hong, the Founder and CEO

• Xiao Hong studied at Huazhong University of Science and Technology in Wuhan. Even as a student, he built tools for WeChat.
• Then, in 2015, he launched a startup and a WeChat management tool called Yiban Assistant.
• Then, in 2019, he created Weiban Assistant, a tool designed for business users of WeChat.

In fact, a firm called Minglue Technology bought that product. So what does this acquisition tell you? Xiao is a product operator. He spots a big platform and builds useful tools on top of it.

Ji Yichao, the Chief Scientist

Ji Yichao, also known as “Peak,” is the technical mind. He left high school at 17 to build Mammoth Browser for the iPhone. Soon after, investor Xu Xiaoping backed him.

• He also made Forbes China’s 30 Under 30 list in 2012 and 2013.
• Then, in late 2022, he joined Xiao to build the company’s next product.

Zhang Tao, the Product Lead

Meanwhile, Zhang Tao joined the founding team as product director. His job was to shape how people actually use the agent.

Why This Mix Matters

Here is a unique point most articles miss. Xiao’s whole career was about building on someone else’s platform. First it was WeChat. Later, he worked with large language models developed by other labs.

As a result, Manus follows the same playbook. It does not train its own base model. Instead, it builds a smart layer on top of the best models available.

That is a key detail if you are wondering what is Manus AI at its core. It is less a new brain and more a new body for existing AI. In other words, the founders bet that the “hands” of AI would matter as much as the “mind.”

Before Manus: Butterfly Effect and Monica

Of course, Manus did not appear out of nowhere. Its parent company is called Butterfly Effect.

A Company Born Before ChatGPT

Xiao founded Butterfly Effect in 2022. Notably, that was about two months before ChatGPT went public. At first, the team worked from offices in Beijing and Wuhan. Even so, they aimed at users outside China, such as those in North America, Japan, and South Korea.

Monica: The First Big Product

Next came Monica in 2023. It started as a browser extension powered by ChatGPT. Over time, it pulled several AI models into one simple window. As a result, users could translate, summarize, and write without jumping between apps.

Monica did well. In fact, ByteDance offered about $30 million to buy the company in 2024. Still, Xiao said no. Reflecting on the past, that decision was highly beneficial.

Early Backers

Meanwhile, funding gradually increased. First, ZhenFund led a seed round in 2023. Then, in late 2024, a Series A brought in Sequoia China (now HSG) and Tencent. These Manus AI investors would stay with the company for years.

The Lesson Hidden in Monica

Here is the insight. Monica taught the team one simple truth. Users will pay for convenience, not for model names. People did not care which model answered. Instead, they cared that the answer came fast and in one place.

Manus takes that same idea much further. Rather than giving you answers, it tries to finish the task. So Monica was not a side project. Instead, it served as a training ground for a larger investment.

How Manus Was Born

Then, in October 2024, the team began work on Manus. Interestingly, they drew ideas from Cursor, the AI coding tool. Cursor showed that AI could act, not just talk. So the founders asked a bold question. What if an agent could work like that for any task, not just code? That question became the seed of the Manus AI agent.

The Launch That Made Manus Go Viral

Next, let’s look at the moment the world noticed.

March 6, 2025

On March 6, 2025, the team launched Manus as an invite-only beta. The demo video passed one million views in about 20 hours. Soon, people called it China’s “next DeepSeek moment.”

What Is Manus AI, in Simple Terms?

So, what is Manus AI? Put simply, it is a general AI agent. You give it a goal, such as “compare five laptops and make a table.” Then it plans the steps, browses the web, writes code if needed, and returns the final result. Meanwhile, it runs on its own cloud computer, so you can close the tab and come back later.

The Invite Code Frenzy

Limited access transformed invite codes into a highly sought-after commodity. In China, some were reportedly resold for 50,000 to 100,000 yuan, roughly $7,000 to $13,800. Even in the US, codes reportedly sold for over $1,000. Clearly, demand was far ahead of supply.

The “Wrapper” Debate

However, not all the buzz was positive. Critics soon asked if Manus was “just a wrapper” around other models. In fact, the team said it used existing models, including Anthropic’s Claude, rather than its own. Still, that answer split opinions.

A Fresh Take on the Launch

Here is a perspective you rarely see. The invite system was not only a marketing trick. Agents use far more computing power per task than chatbots. So the team had to limit users at first. In other words, a real limit turned into a hype engine.

For you, the takeaway is clear. If your product is costly to run, a waitlist can protect your budget. At the same time, it can also build demand. Just be honest about why the wait exists, so trust grows along with the hype.

Funding, Fast Growth, and the Move to Singapore

After the launch, progress accelerated rapidly. Let’s walk through the key steps.

The Benchmark Round

First, in April 2025, Manus raised about $75 million in a Series B. Silicon Valley firm Benchmark led the round. As a result, the round valued the company at about $500 million. Tencent, HSG, and ZhenFund also joined.

However, the deal drew attention in Washington. Reports said the US Treasury looked into it under new rules on investing in Chinese tech. Benchmark’s lawyers argued the rules did not apply. After all, Manus did not train its own base models, and its parent was registered in the Cayman Islands.

The Big Move to Singapore

Next, in mid-2025, the company moved its headquarters to Singapore. The three co-founders moved too. Of about 120 staff in China, around 40 key engineers relocated. Sadly, the rest lost their jobs.

On top of that, Manus blocked access in mainland China. It also shelved a planned Chinese version built with Alibaba’s Qwen team.

Record Revenue Growth

Meanwhile, revenue soared. By August 2025, the run rate hit about $90 million. Then, in October, the team released Manus 1.5 with faster, more reliable results. By December 2025, Manus passed $100 million in annual recurring revenue.

That came only about nine months after launch. Later, reports put the run rate near $500 million by June 2026.

Our Take on the Singapore Move

Most articles treat the move as pure politics. But there is also a business angle. Manus needed top US models and global payment systems. Those were difficult to use from within China. So the move was as much about market access as about safety.

What You Can Learn

If you run a startup, ask yourself three questions:

1. Where are your paying users? Build your base close to them.
2. Which tools do you depend on? Make sure you can legally use them where you operate.
3. Who owns your company on paper? Your legal structure can shape your future deals.

The Meta Deal and Its Dramatic Reversal

Now we reach the twist that made global headlines.

Meta Steps In

In December 2025, Meta announced it would buy Manus. Reports put the deal at over $2 billion. Meta said it would continue selling Manus as a product. It also planned to bring the tech into tools like Meta’s AI. For many, this was the perfect startup ending.

China Pushes Back

Yet the story did not end there. In January 2026, China’s Ministry of Commerce opened a review of the deal. Soon after, reports said Chinese authorities had barred some Manus executives from leaving the country.

Then, on April 27, 2026, China’s top planning agency blocked the deal. It ordered both sides to withdraw it. Meta said the deal obeyed the law. Even so, Meta cut ties with Manus in June 2026.

Manus Goes Solo Again

Thereafter, in August 2026, Manus said it would run as an independent company again. As part of the split, it said it would delete some user data. Then, in October 2026, Manus raised more than $500 million.

Boyu Capital and IDG Capital led the round. Meanwhile, Tencent, HSG, and ZhenFund returned too. There is even talk of a Hong Kong listing, though not before 2027.

Why This Matters for Founders

Here is a lesson that goes beyond Manus. Moving your office does not erase your roots. Even after the shift to Singapore, the founders’ links to China still shaped the deal. In short, where you come from can matter as much as where you are.

Your Action Steps

If you hope to sell your startup one day, plan early:

• Map your risks. List every country tied to your founders, staff, and code.
• Talk to lawyers before deals. Ask about national security reviews in each country.
• Keep a Plan B. Manus survived because its product and revenue stayed strong on their own.

Five Lessons Founders Can Take From the Manus AI Story

Finally, let’s turn the story into simple steps you can use.

Lesson 1: Build on Platforms, Then Go Deeper

• First, notice the pattern. Xiao built on WeChat, then on AI models. So, start with what exists.
• Then add a layer that saves people real time. For example, ask which daily task your users hate most.
• Next, build a tool that removes that pain on top of an app they already use.

Lesson 2: Say No to Early Offers When the Vision Is Bigger

Next, remember the $30 million ByteDance offer. Xiao turned it down. That choice later led to a deal worth over $2 billion. Of course, such a decision is not always right. Still, know your value before you sell.

Lesson 3: Show, Don’t Tell

Furthermore, Manus won attention with a demo, not a press release. People saw the agent work. Therefore, if your product does something visual, film it. Keep the clip short, and show one real task from start to finish.

Lesson 4: Turn Limits Into Momentum

Likewise, the invite system turned a computing limit into hype. Therefore, look for ways your constraints can build demand.

Lesson 5: Plan for Politics Early

Lastly, the Meta saga shows that geopolitics can sink a deal. So, think about legal and regional risk from day one, not on deal day. In fact, a short yearly check with a good lawyer costs far less than a failed sale.

Why These Lessons Stand Out

Here is the bigger picture. The Manus startup story is not about one lucky launch. Rather, it shows a team that kept stacking up small victories. A WeChat tool led to Monica. Then Monica led to Manus. In the end, each step paid for the next one. You can follow that same path.

Conclusion

So, who built Manus AI? At its heart, it was a small team with a clear pattern. Xiao Hong brought product sense and years of building on big platforms. Ji Yichao brought exceptional technical skill. Meanwhile, Zhang Tao shaped the product people use every day.

Together, they turned a browser extension called Monica into one of the fastest-growing AI startups in the world. Along the way, they went viral, raised money from Benchmark, and moved to Singapore.

Then they agreed to a $2 billion sale to Meta, only to see China block it. Even so, the company bounced back with a new $500 million round in October 2026.

The main message is simple. Great products often grow from smaller ones. In addition, the best founders combine bold vision with careful planning. Manus shows both the rewards and the risks of building across borders.

Now it is your turn. If you are building a startup, look at your own “Monica.” Which small product could teach you what users truly want? Furthermore, check your legal and regional risks before your big moment arrives.

Want more stories like these? Explore more founder journeys on Businesstories.com. You can also share this listicle with a friend who is building the next big AI tool.

FAQs About Who Built Manus AI

1. Who are the founders behind Manus AI, and what are their roles today?

Manus AI was built by Xiao Hong, the founder and CEO, and Ji Yichao, the chief scientist. Zhang Tao joined the founding team as product director. All three moved to Singapore in 2025.

2. What company owns the Manus AI agent now after the failed Meta deal?

Butterfly Effect, the startup founded by Xiao Hong in 2022, now owns the Manus AI agent. After China blocked the Meta deal, Manus became independent again in August 2026 and raised over $500 million.

3. Why did the Manus AI startup move its main headquarters to Singapore?

The Manus AI startup moved to Singapore in mid-2025 after its benchmark round. The move helped it reach global users, use leading US models, and reduce regulatory pressure linked to China.

4. How did Monica help the founders create the Manus AI agent later?

Monica was an AI browser extension that brought many models into one place. It taught the founders that users value convenience. Manus took that idea further by finishing whole tasks for users.

5. Why did China block Meta’s $2 billion purchase of Manus AI in 2026?

China reviewed the deal amid concern over US investment in Chinese-founded AI startups. In April 2026, its planning agency blocked the purchase and ordered both sides to withdraw, so Meta later cut ties.