Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Manus helped turn AI agents into China’s next major technology race, but it did not create the market alone. Its March 2025 launch gave consumers, investors, and technology companies a highly visible example of an AI system that could plan and execute multistep tasks. That made “agents” more tangible—and more investable—than another chatbot demonstration.
The boom also depended on forces already in motion: DeepSeek’s momentum, cheaper domestic models, China’s large digital platforms, cloud infrastructure, venture funding, and government pressure to commercialize artificial intelligence. Manus was best understood as a catalyst and category-defining signal, not the sole cause.
As an Amazon Associate I earn from qualifying purchases.
What Manus changed
A chatbot generally responds to a prompt. An AI agent is designed to pursue an objective by planning several steps and using tools such as web browsers, code environments, files, APIs, or business software.
Free tools Windows power users keep installed
One-click scans. No signup required.
Manus was marketed as a general-purpose agent for tasks including research, coding, market analysis, website creation, presentations, and document production. Its significance was not simply that it generated text. The product promised to turn an instruction into a completed workflow.
#1 Best Overall
That distinction matters. Calling a product an “agent” does not prove that it can operate reliably without supervision. Academic descriptions characterize systems such as Manus as combinations of language-model reasoning, planning, tool use, and task execution, while public evidence about their reliability across difficult, open-ended work remains limited. Research on Manus’s architecture should not be confused with an independent guarantee of performance.
Why the March 2025 launch attracted attention
Manus launched publicly on March 6, 2025, initially through an invite-based rollout. Reports described intense interest in invitation codes and heavy traffic around registration. Its demonstrations emphasized autonomous completion: users assigned a project rather than asking a single question.
The timing amplified the reaction. DeepSeek had already encouraged Chinese users, investors, and policymakers to believe that a relatively low-cost AI product could have global impact. Manus shifted that excitement from the model layer to the application layer. The question became less “Which company has the strongest model?” and more “Can an AI system actually do useful work?”
Contemporary coverage compared the launch with a possible second DeepSeek moment, although that comparison was a media narrative rather than proof of equivalent technical or commercial importance. China.org.cn’s launch coverage documents the timing and early attention.
What evidence supports an AI-agent boom?
There is clear evidence of a surge in launches, corporate competition, investment interest, and public experimentation. But these are not interchangeable measures. A long list of products proves competitive activity; it does not prove durable revenue, high retention, or reliable autonomous work.
1. A wave of competing products
Following Manus, Chinese or Chinese-founded companies promoted a range of general-purpose and specialized agents. Names appearing in contemporary coverage included Genspark, Flowith, Fellou, Alibaba’s Quark, Zhipu’s AutoGLM-related products, and agent offerings from Tencent, ByteDance, Alibaba, and Baidu.
These products should not be treated as identical competitors. Some aimed to reproduce Manus’s open-ended workflow, while others were assistants embedded in cloud, search, office, commerce, or enterprise software. Product names, availability, and capabilities changed quickly, so a launch announcement is safer evidence than assuming every listed product remains active or comparable.
Recommended Free Tools
2. Big technology companies entered the race
Large platforms had advantages that startups could not easily duplicate: cloud capacity, distribution, existing enterprise customers, data connections, and permission systems.
At the World Artificial Intelligence Conference in Shanghai in July 2025, Tencent’s cloud unit reportedly showcased about a dozen AI agents. Alibaba promoted agent infrastructure, Qwen-powered tools, Quark, and Wukong for enterprise use. Baidu, ByteDance, and Zhipu also developed agent products around their own models and software ecosystems. The South China Morning Post’s account of the 2025 competition illustrates how agents moved from startup novelty to strategic battleground.
3. Investors saw a new application layer
Manus made a commercially attractive proposition visible: an AI company might charge for completed work rather than merely for access to a chatbot or model.
That idea attracted attention, but the business case remained difficult. General-purpose agents can consume substantial inference and tool-use resources. Users may switch easily among products. Platform companies can bundle similar features with cloud or productivity services. Enterprise buyers also demand security, uptime, support, compliance, and auditability—requirements that are much harder than producing a convincing demo.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsManus said it had raised $75 million led by Benchmark; that figure is company-reported. Its own funding and revenue announcement should therefore be read as a corporate disclosure, not an audited measure of market-wide profitability.
Who followed Manus?
General-purpose startup agents
Genspark, Flowith, Fellou, and Manus pursued the broadest version of the promise: give the system an open-ended assignment and let it coordinate research, browsing, writing, coding, presentations, or other actions.
Genspark, for example, has positioned its “Super Agent” around workflows such as phone calls, presentations, and video creation. Vendor demonstrations and customer case studies can show intended use cases, but they are not independent benchmarks of accuracy or successful completion rates. OpenAI’s customer account of Genspark is useful for understanding the product’s positioning, not for proving universal autonomy.
Platform-integrated agents
Alibaba, Tencent, ByteDance, Baidu, and Zhipu could embed agent functions into products that already have users, cloud accounts, data connections, and enterprise relationships.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →- Alibaba: Qwen models, Quark, Alibaba Cloud tooling, and Wukong-style enterprise-agent services.
- Tencent: Cloud-hosted agents and integrations with its broader software ecosystem.
- ByteDance: Doubao and related AI applications.
- Baidu: ERNIE-based products and enterprise AI tools.
- Zhipu: AutoGLM-related agent products and developer services.
Platform distribution can lower customer-acquisition costs and simplify deployment. The trade-off is that platform agents may be less open-ended, more tightly governed, or more dependent on a vendor’s ecosystem.
Open-source and developer-led agents
In 2026, the center of attention broadened beyond polished, invite-based products. The OpenClaw craze involved users deploying agents on rented cloud servers, developers connecting them to domestic models such as Qwen and Kimi, cloud providers promoting hosted versions, and service providers helping people install and configure systems.
This was a different adoption model from Manus. Instead of one company providing a finished product, users and developers assembled agents from software, models, hosting, integrations, and services. Reports of queues for installation help and widespread experimentation show strong interest, but they do not establish a precise nationwide user count or prove that Manus directly caused the later wave. Associated Press coverage and Wired’s reporting describe both the enthusiasm and the reliability gap.
Rank #4
Why China was fertile ground for agent adoption
Manus benefited from a market that was already prepared to experiment with AI applications.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- Digital consumer and business ecosystems: China’s large online platforms and super-app-style services create many potential integrations for search, payments, commerce, communications, and administration.
- Domestic models: Qwen, DeepSeek, Kimi, Doubao, GLM, and other systems provide local alternatives and can lower model-access costs.
- Cloud infrastructure: Alibaba Cloud, Tencent Cloud, Baidu, and other providers can host agents and sell the required computing capacity.
- Competitive pressure: Companies and workers have strong incentives to automate research, marketing, coding, customer service, and routine administration.
- Fast product cycles: Large internet firms and startups can launch, copy, and iterate quickly.
- Commercialization priorities: Government and industrial demand can accelerate pilots and procurement.
China is not a frictionless test market. Security rules, privacy requirements, censorship controls, data-localization concerns, model-access restrictions, and enterprise governance can complicate deployment. An agent that works in a public demonstration may require substantial modification before it can access sensitive company systems.
What counts as a real AI agent?
The label is being applied unevenly. A practical taxonomy helps separate marketing language from capability:
- Chat assistant: Generates answers or content in response to a prompt.
- Tool-using assistant: Calls search, code, browsing, or business tools, usually with frequent user direction.
- Workflow agent: Executes a defined sequence such as lead qualification or report generation.
- Open-ended general agent: Plans and adapts across multiple steps with relatively little supervision.
- Agent platform: Lets developers or enterprises build, host, monitor, and govern agents.
Many products marketed as agents are likely to fall somewhere between categories two and four. That is still useful, but it is not the same as a fully autonomous digital employee. Industry commentary has warned that inconsistent use of the term can make products appear more capable than they are. Rest of World’s reporting discusses this variation in China’s agent market.
The demo-to-deployment gap
Agents can offer real advantages:
- Automating multistep knowledge work.
- Turning natural-language instructions into actions.
- Combining research, code, files, and business tools.
- Allowing small teams to handle work that previously required several specialists.
- Providing customized workflows when open-source systems are connected to domestic models.
They also create serious operational risks:
- Hallucinated facts, fabricated sources, or incorrect calculations.
- Wrong actions caused by ambiguous instructions.
- Fragile browser and API integrations.
- High and unpredictable costs on long tasks.
- Lost context across extended workflows.
- Security exposure when agents receive access to email, files, payments, or internal systems.
- Weak audit trails and unclear accountability.
- Privacy, data-residency, and regulatory problems.
Human review remains essential for legal, financial, medical, employment, and other high-stakes decisions. A useful evaluation should ask how many tasks are completed without correction, what each successful workflow costs, how often users return after the first task, and whether customers pay recurring fees. Viral demonstrations answer none of those questions.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhy many agent startups may struggle
A startup can show impressive performance and still lose to a platform company. Distribution is expensive, inference costs can overwhelm subscription revenue, and users may move between agents with little switching cost. Cloud providers can bundle agents with infrastructure, while chat and productivity platforms can add similar features to products customers already use.
Best Value
Enterprise buyers further raise the bar. They want role-based access, logging, security reviews, predictable spending, service-level commitments, and support. General-purpose autonomy is difficult to differentiate once competitors copy the interface and basic workflow.
That makes business durability more important than launch count. For investors and business leaders, the key questions are active usage, retention, paid customers, cost per successful workflow, human-correction rates, and audited revenue—not merely whether a product went viral.
Manus’s timeline and geopolitical afterlife
- March 6, 2025: Manus launched publicly through an invite-based rollout. Contemporary coverage reported the launch and early attention.
- Spring 2025: The product gained international attention and was compared with DeepSeek. Axios described the early reaction.
- Mid-2025: Reporting described a relocation of key operations or leadership from China to Singapore and a reduced mainland presence. The exact corporate and operational structure should not be simplified to a change of headquarters.
- December 2025: Meta announced an agreement to acquire Manus. The announcement did not initially disclose a precise purchase price; outside reports later estimated a value of roughly $2 billion to $3 billion.
- January–April 2026: Chinese authorities reviewed the transaction, with reporting also describing restrictions involving the founders’ ability to leave China.
- April 27, 2026: AP, Bloomberg, Axios, and TechCrunch reported that Chinese authorities blocked or ordered the transaction unwound. A regulatory prohibition, an order to unwind, operational separation, and a legally completed rescission are not necessarily the same event.
- June–July 2026: Reports described Manus separating from Meta and possible interest from Chinese investors or Tencent in reacquiring the company. Those reports should not be treated as confirmation of a completed transaction without an official announcement.
AP’s reporting on the transaction, together with accounts from The Washington Post and TechCrunch, shows why the company’s ownership became a strategic issue.
The episode demonstrates that “Chinese-founded,” “Chinese-based,” and “Singapore-based” are not interchangeable descriptions. Founders, employees, investors, intellectual property, source code, data, incorporation, and operating headquarters can all matter independently. Relocating a company does not automatically remove regulatory claims over technology, assets, or personnel.
What the Manus story means for businesses and investors
For a business considering an agent, the practical choice is usually not between “AI” and “no AI.” It is between levels of control and risk:
- Hosted general-purpose agent: Easiest to try, but typically offers less control over data, model routing, costs, and availability.
- Cloud enterprise platform: Better suited to managed deployment, integrations, and governance, but may require more setup and vendor commitment. Alibaba Cloud and Tencent Cloud are examples of providers serving this segment.
- Self-hosted or open-source system: Offers customization and the ability to connect domestic models, but shifts security, monitoring, credential management, and maintenance to the buyer.
Before deployment, organizations should require sandboxing, approval gates for consequential actions, spend limits, access controls, logging, data-handling documentation, and a fallback when the agent fails. They should also calculate the total cost of model calls, cloud hosting, storage, monitoring, integration work, and human review.
Verdict: catalyst, not creator
Manus did not create Chinese agentic AI from nothing. China already had the models, platforms, cloud providers, talent, capital, and policy incentives needed for an agent market.
What Manus did was make the category vivid. Its invite-only launch and task-completion demonstrations changed the reference point from “Which chatbot answers best?” to “Can an AI system complete a project?” That attention effect helped startups attract interest and pushed major platforms to treat agents as a strategic product category.
The later OpenClaw wave shows that the market’s evolution did not depend on Manus alone. Adoption broadened into open-source software, domestic models, cloud hosting, local services, and developer experimentation. The enduring winners will be determined less by viral demos than by reliable task completion, manageable costs, secure deployment, recurring revenue, and the ability to operate within China’s increasingly strategic AI rules.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
