AIWiki
Malaysia
Back to all articles
Tools & Platformsmanusai agentsagentic ai

Manus (AI Agent)

4 min readUpdated September 2026
Manus
Type
Autonomous AI agent
Developer
Butterfly Effect Pte. Ltd.
Initial release
6 March 2025 (invitation beta)
Latest version
Manus 1.6 (December 2025)
Licence
Proprietary
Website
manus.im
Manus is an autonomous artificial intelligence agent developed by Butterfly Effect, a company founded in China and headquartered in Singapore. Launched in March 2025, it carries out multi-step knowledge tasks — market research, data analysis, document creation and web application development — with limited supervision, and became one of the most closely watched products of the industry's "agentic AI" wave.[1]

History

Butterfly Effect was founded in 2022 by Xiao Hong, two months before the launch of ChatGPT. The company, with offices in Beijing and Wuhan, first released Monica, a browser extension that aggregated several large language models, and targeted markets outside China — primarily North America, Japan and South Korea.[1][2]

In 2024, ByteDance reportedly offered about US$30 million to acquire the company; the offer was declined. Development of Manus began in October 2024, drawing inspiration from the AI coding tool Cursor.[1]

Manus launched in invitation-only beta on 6 March 2025. A demonstration video showing the agent autonomously screening résumés and analysing stocks drew more than a million views within twenty hours, and invitation codes were resold on secondary markets for prices reported from hundreds to thousands of US dollars. In April 2025 the company raised approximately US$75 million in a funding round led by Benchmark at a reported valuation of about US$500 million; the investment briefly drew scrutiny from United States authorities under rules governing outbound investment in Chinese technology, an issue that faded after the company moved its base.[1]

In mid-2025, Butterfly Effect relocated to Singapore, where a new entity took over operation of the product outside China. Manus 1.5, released in October 2025, introduced full-stack web application generation — producing working prototypes with backend logic, databases and authentication — alongside reported improvements in speed and reliability; Manus 1.6 followed in December 2025.[1][3]

In December 2025, Meta agreed to acquire Manus in a deal valued at more than US$2 billion as part of its push into AI agents. In April 2026, Chinese regulators blocked the transaction and ordered Meta to unwind it, illustrating the geopolitical pressures on AI companies operating between the United States and China — a sequence that reporters later described as a "cautionary tale" for the sector.[4][5][6]

Technology

Manus belongs to the class of general AI agents: systems that plan, execute and verify multi-step work rather than only answering questions. A user gives a goal; Manus breaks it into steps, operates tools such as a web browser, code interpreter and file system inside a cloud sandbox, and returns finished output — a report, a dataset, a slide deck or a small web application. Tasks run asynchronously, so work can continue while the user is offline.[1][3]

Unlike model developers such as OpenAI and Anthropic, Butterfly Effect does not train its own foundation models; Manus orchestrates third-party models through its own agent framework, tooling and user experience. The company reported state-of-the-art scores on the GAIA benchmark for general assistants at launch, and has continued to emphasise execution speed and reliability in later versions.[1][3]

Applications and impact

Manus is used for market and competitor research, financial analysis, document and slide production, travel planning and rapid prototyping of web applications. Its launch intensified competition in AI agents among OpenAI, Google, Anthropic and Chinese developers, and its journey — rapid growth, a relocation to Singapore and a blocked acquisition — has been widely cited as a case study in how geopolitics now shapes the AI industry.[1][6]

>See Also

🇲🇾Malaysian Context

🇲🇾 Autonomous agents such as Manus are part of the toolkit that Malaysian knowledge workers, agencies and small businesses are beginning to trial, following widespread adoption of assistants such as ChatGPT. The tools appeal to lean teams that need research and documentation capacity without hiring, and Manus's English-language workflows and Singapore base make it accessible to businesses across ASEAN.

Adoption brings governance questions. Giving an agent access to email, documents and online accounts raises data-protection and cybersecurity risks, and organisations are advised to scope agent permissions carefully and avoid entering sensitive personal data into external services — considerations that sit within Malaysia's Personal Data Protection Act 2010 and the remit of agencies such as NACSA, the National Cyber Security Agency. Pricing in US dollars is a further consideration for smaller Malaysian firms, while MDEC's digital-talent programmes increasingly include practical training in building and supervising AI agent workflows.

References

  1. ↑Wikipedia. (2026). Manus (AI agent). https://en.wikipedia.org/wiki/Manus_(AI_agent)
  2. ↑South China Morning Post. (2025). Chinese AI agent Manus transcends chatbots, start-up's founder says. https://www.scmp.com/tech/tech-trends/article/3301864/chinese-ai-agent-manus-transcends-chatbots-founder-butterfly-effect-says
  3. ↑Manus. (2025). Introducing Manus 1.5. https://manus.im/blog/manus-1.5-release
  4. ↑Taipei Times. (2025). Meta to acquire Manus in US$2bn bet on AI. https://www.taipeitimes.com/News/biz/archives/2025/12/31/2003849775
  5. ↑Reuters. (2026). China orders Meta to unwind $2 billion purchase of AI startup Manus. https://www.reuters.com/world/asia-pacific/china-blocks-foreign-acquisition-ai-startup-manus-2026-04-27/
  6. ↑France 24. (2026). Caught between great powers: the cautionary tale of Manus. https://www.france24.com/en/live-news/20260811-caught-between-great-powers-the-cautionary-tale-of-manus