Manus
Meta reportedly agreed to acquire it for $2 billion — then in April 2026, China's National Development and Reform Commission ordered the deal unwound on foreign-investment security grounds. Manus remains independent.
What is Manus?
Manus is a general-purpose autonomous AI agent built by Butterfly Effect, an AI startup founded in China and headquartered in Singapore since mid-2025. It goes well beyond chatbot-style question answering: describe a goal, and Manus independently plans multi-step tasks, browses the web in real time inside its own virtual computer, writes and executes code, analyzes data, and delivers a finished result from a single prompt — booking flights, filling out spreadsheets, and compiling research without any human hand-holding at each step. Its launch story was genuinely explosive: a viral, invite-only demo in March 2025 sent its waitlist past six figures within 48 hours, with the obvious question on everyone's mind being whether this was the agent that would finally replace half a typical SaaS stack. Sixteen months later, the honest answer is more nuanced than that initial hype suggested — real-world testing found execution still uneven as of mid-2026, with tasks occasionally failing mid-stream, and 2026 feature additions (a Web App Builder, Slack/WhatsApp/Telegram integrations, the Manus 1.6 model family, Scheduled Tasks 2.0, Gmail and Google Calendar connectors, and "Wide Research" running parallel sub-agents) reflect continued, active investment in closing that gap.
The genuinely wild, current story worth knowing about Manus's 2026 corporate journey: Meta reportedly agreed to acquire the company for roughly $2 billion, but in April 2026, China's National Development and Reform Commission ordered the transaction unwound entirely on foreign-investment security grounds — meaning Manus remains independent rather than becoming part of Meta's portfolio, a real, significant geopolitical twist in an AI acquisition story. It's honest to be precise about where Manus's real strength lies, and where it doesn't: one detailed hands-on review notes plainly that its app-building feature generates functional prototypes, not production code — database schemas are naive, error handling is minimal, and security best practices are largely absent, meaning anyone needing AI-generated code they can actually ship is better served by a tool like Devin, Replit Agent, or Claude Code. Manus's genuine strength lies in orchestrating non-coding workflows — research-heavy tasks visiting many websites and compiling structured data specifically, where the time savings are real and meaningful.
The Meta acquisition and its reversal by Chinese regulators are drawn directly from an independent, dated 2026 review (Taskade), corroborated by a separate pricing source (No Code MBA). The honest scope clarification on prototypes versus production code is drawn from a detailed hands-on independent review (ToolChase). Credit expiration policy details are drawn from an independent TCO analysis (CheckThat.ai).
Key features
Virtual computer environment
Browses the web, writes code, and manipulates files inside its own workspace.
Wide Research
Runs parallel sub-agents to compile research from many sources simultaneously.
Web App Builder
Generates functional prototypes and simple deployed websites from a prompt.
Scheduled Tasks 2.0
Automates recurring agent runs on a schedule you define.
Slack, WhatsApp, Telegram, Gmail, Calendar
A genuinely broad set of 2026 connector additions.
Manus 1.6 model family
Lite, standard, and Max variants tuned for different task complexity.
Pricing
Free
- Enough to run roughly 3-5 simple tasks per day
- Chat Mode and Manus 1.6 Lite in Agent Mode included
- Daily credits expire after just 24 hours if unused
Pro
- Adds Manus 1.6 and 1.6 Max, Wide Research, website deployment
- 20 concurrent and scheduled tasks on entry Pro tiers
- Only purchased add-on credits carry forward, and only while subscribed
Team / Enterprise
- SSO, usage analytics, access controls, shared credit pooling
- Enterprise requires direct contact and consultative pricing negotiation
- Master Services Agreement is not publicly accessible
Idle time, storage, and completed work do not consume credits — only active task execution does. Annual billing saves 17% across all paid tiers. Prices reflect Manus's published pricing as of July 2026.
Available models
Integrations & platforms
Pros, cons & best for
Pros
- Genuinely capable, autonomous research and multi-step task execution
- Real, ongoing feature investment (Wide Research, new connectors, model upgrades)
- Idle time and completed work don't consume credits, a fair usage policy
Cons
- Execution remains genuinely uneven, with tasks occasionally failing mid-stream
- Generated code is prototype-quality, not production-ready
- Daily credits expire in 24 hours, and monthly credits don't roll over either
Best for
- Individual research-heavy tasks visiting many websites and compiling data
- Solo users wanting a general-purpose autonomous agent, not a team platform
- Not the pick for shippable production code — try Devin or Claude Code instead
Take a look inside
Alternatives
For an open-source single-agent framework or multi-agent orchestration instead:
Our verdict
Manus captured real, explosive attention for good reason — a general-purpose agent that plans, browses, codes, and delivers finished work from a single prompt is a genuinely compelling vision, and features like Wide Research reflect real, continued investment in making that vision more reliable. The honest, current reality sixteen months in: execution remains genuinely uneven, and the app-building feature specifically produces prototypes rather than shippable production code, so setting expectations accurately matters before relying on it for anything client-facing. The Meta acquisition saga — a reported $2 billion deal unwound by Chinese regulators in April 2026 — is a genuinely remarkable, current story worth knowing, even if it doesn't change what the product itself can do today. For individual research-heavy work and general-purpose task automation, Manus remains a genuinely capable, closely-watched choice; for coding you intend to ship, or team collaboration features, look elsewhere first.
FAQ
Did Meta acquire Manus?
Meta reportedly agreed to acquire Manus for around $2 billion, but in April 2026, China's National Development and Reform Commission ordered the transaction unwound on foreign-investment security grounds, meaning Manus remains an independent company.
Can Manus generate production-ready code?
No, its app-building feature generates functional prototypes rather than production code — database schemas are naive and security best practices are largely absent, so tools like Devin, Replit Agent, or Claude Code are better suited for shippable code.
Do unused Manus credits carry over to the next period?
No — daily refresh credits expire after just 24 hours, and monthly subscription credits also don't roll over; only separately purchased add-on credits carry forward, and only while your paid subscription remains active.
What is Manus genuinely good at?
Orchestrating non-coding workflows specifically — research-heavy tasks that involve visiting many websites and compiling structured data are where it delivers real, meaningful time savings.
Is Manus reliable for production use?
Independent testing as of mid-2026 found execution genuinely uneven, with tasks occasionally failing mid-stream — worth testing thoroughly on your specific use case before depending on it for critical workflows.
Does Manus support team collaboration?
Team and Enterprise plans exist with SSO, shared credit pooling, and access controls, though some independent reviewers note it originated and still performs best as a solo-use, general-purpose agent rather than a fully collaborative team platform.