New Manus accounts start with 1,000 free credits and a 300-credit daily refresh, though the amount can vary by country. Accounts created from a US IP address and from the UK both opened with the full 1,000, and that bonus sets how far the free plan stretches.
Manus AI is a general-purpose autonomous agent. You describe an outcome, and it plans the steps and runs them in its own cloud environment. It then hands back a finished research brief, website, video, or workflow. It also bills per credit, and that detail shapes everything below.
Hands-on testing here stayed on the free plan. The sections on paid features, pricing and security draw on Manus’s documentation, security researchers’ reports and independent coverage.
| Quick facts (November 2026) | |
|---|---|
| Maker | Butterfly Effect, Singapore-based |
| Current version | Manus 2.0, released September 28, 2026 |
| Free plan | 1,000 one-time credits for new accounts (the amount can vary by country), plus 300 daily credits on the Manus 2.0 Lite agent |
| Paid plans | From $20 a month for 4,000 credits |
| Best for | First drafts you can verify |
| Main risk | Connected accounts and unreviewed output |
What the free plan gave in testing

Manus’s help center promises 1,000 credits at registration, plus 300 daily credits. Accounts created from a US IP address and from the UK both received the full bonus.
| Account | Free credits | Daily refresh |
|---|---|---|
| Created from a US IP address | 1,000 (listed as “Bonus for new users”) | 300 |
| Created in the UK | 1,000 | 300 |
| Created from Pakistan | 0 | 300 (290 after two tasks) |
Manus doesn’t publish a country-by-country list, so the bonus can differ by region.
Spending stayed small. A one-line greeting cost 2 credits in one second. A short Computer Use test that navigated public pages cost 8 credits in 57 seconds. Together, the two tasks pulled the daily balance from 300 to 290. Both ran on the Manus 2.0 Lite agent that the free plan uses. Neither number says much about complex work.

Who owns Manus now?
Manus is independent again. Meta announced the purchase in December 2025, and China’s state planner ordered the deal reversed in April 2026. Manus said it had formally resumed independent operations on September 1, 2026, with its founding team in charge.
In October, parent company Butterfly Effect announced a round of more than $500 million. Boyu Capital and IDG Capital led it, and Tencent, HSG, and ZhenFund joined. Manus hasn’t confirmed a valuation, though media reports put it near $4 billion, and its statement didn’t spell out who owns what.
The split cost some users data. An August 11 notice told affected accounts to back up anything created on or after December 29, 2025, because Manus would delete that data on August 23 and 24.
What Manus 2.0 adds
Manus launched in March 2025 as a research-and-execution agent. Version 2.0 widened the pitch in four directions, according to the company’s launch post.

- Everyday work. Studio, the renamed desktop app, pulls documents, code, websites, video and games into one workspace. Paid plans add Wide Research, which splits a research job across parallel sub-agents.
- Building things. Manus’s documentation says a built site can include a backend, database, user login, and Stripe payments, with hosting handled for you.
- Video and games. The Video Editor lays out separate clips, music and text on a timeline, so swapping a song doesn’t mean regenerating the whole video. Game Dev builds playable games, multiplayer ones included, and persistent servers run on a paid Cloud Computer.
- Work that runs without you. Automations start a task when something happens in a connected app: Gmail, Outlook, Google Calendar, GitHub, Shopify, Slack, Notion, RSS or a webhook. Computer Use lets Manus operate apps and the browser on your machine through the Studio desktop app, with permission modes you pick and an operating-system prompt the first time it touches a folder.
Manus also connects to Zapier. A separate app called Cue, launched alongside 2.0, gives each agent its own email address, phone number, wallet, and computer. Cue began as free, invite-code early access, and its App Store listing still asks for a code.
The headline engineering change is Cascade, a rebuilt agent harness that keeps a project light and calls in specialized capabilities only when a step needs them. Manus reports that in one test configuration Cascade used 23.2% fewer tokens, finished tasks 28.2% faster and cost 32% less than the previous system. Those are company-reported figures from a single setup, and no independent replication turned up.
Where capability and reliability split
A polished output can still hide mistakes. Research needs its sources and conclusions verified. Websites need hands-on testing, since a page that renders can still break on login or payment. Automations carry the biggest stakes, and agent automation fails silently: a script throws an error, while an agent hands back a plausible answer that happens to be wrong. Anything that spends money, touches customer records, or messages a real person needs a human approval step.
What Manus costs

These prices reflect Manus’s published plans in November 2026. They can change, and the pricing page at manus.im/pricing carries the current numbers.
| Plan | Price | Credits |
|---|---|---|
| Free | $0 | 300 daily, plus a one-time 1,000 bonus (varies by country) |
| Pro | $20 | 4,000 a month |
| Pro | $40 | 8,000 a month |
| Pro | $200 | 40,000 a month |
| Team | $20 per seat a month, or $200 a year | 4,000 per seat, pooled |
Every paid plan adds 300 daily refresh credits, 20 concurrent tasks, and 20 scheduled tasks. Annual billing brings the three Pro tiers to $17, $34, and $167 a month. Monthly plan credits don’t roll over. All three Pro tiers price credits at $5 per 1,000, so a bigger tier buys volume rather than a discount. Team adds admin controls, audit logs, and SOC 2 Type II and ISO 27001 compliance.

Manus can’t quote a task’s cost in advance. Its credits page offers three worked examples: 200 credits for a 15-minute data chart, 360 for a 25-minute wedding-invitation page, and 900 for an 80-minute app. At the Pro rate, that works out to roughly $1 to $4.50. Long or multi-step jobs can run far higher, and the free-plan numbers above predict little about them. Credits drain only while a task runs, and Manus refunds credits when a task fails because of a fault on its side.

Always-on projects cost extra, because a Cloud Computer bills separately from the credit plan. Linux Standard (4 GB RAM) runs $30 a month and Linux Advanced (8 GB) $50. The plan screen on November 16 also listed Windows Server from $50, with the Advanced spec at $90, a high-spec 3D rendering machine at $200, and macOS from $200, which had shown as sold out on an earlier visit. Bandwidth beyond the included amount costs $0.15 per GB.
Phone calls cost $9.99 a month plus credits
Settings includes a Phone tab that assigns a virtual number to your agent so it can place calls and send messages. Setup opens with three acknowledgments. AI delivers the calls and messages on your behalf. You need a legitimate reason to contact each recipient and must follow applicable laws. Communications may be recorded and stored for compliance and verification.
The purchase screen is region-based and notes that numbers are available in selected regions only. On the US account, a United States (+1) number costs $9.99 a month, calls draw 0.5 credits per minute, and SMS is unavailable because US numbers support voice only.

On the UK account, a United Kingdom (+44) number also costs $9.99 a month, but calls draw 3.172 credits per minute, and SMS is available, charged in credits. At the Pro rate of $5 per 1,000 credits, a 10-minute US call costs about 2.5 cents, and a 10-minute UK call about 16 cents, so the monthly fee stays the main bill for light use. Only the setup and purchase screens were reviewed here, which leaves call quality untested.

The legal weight sits with the account holder. In February 2024, the FCC ruled that AI-generated voices count as “artificial” voices under the Telephone Consumer Protection Act, which requires prior express consent for such calls unless an exemption applies. Anyone planning outbound AI calls to leads or customers should get legal advice for their own situation first.
Is it safe to connect your accounts?
It depends on what gets connected. Public research and throwaway prototypes carry little risk. A work inbox, customer data, or a code repository with live credentials carries much more.
Two documented attacks show why. In 2025, researcher Johann Rehberger showed that a poisoned document could push Manus into exposing its internal VS Code server to the internet and sharing the password with an attacker.

In September 2026, Dark Reading reported a Salt Labs finding: one email hiding obfuscated code could make Manus run attacker code once a user asked it to go through their inbox. Manus’s built-in warning did fire, but only after the code had executed. Salt’s October 1 write-up says connected email, cloud storage, and code repositories were within reach. Salt says the flaw went through Meta’s bug bounty program and is resolved, and Dark Reading reports it was patched.
The pattern outlasts the patch: an agent reads untrusted content while holding your credentials. Event-triggered Automations could remove the one step the attack needed from a victim, since an email can now start a run on its own. Giving Manus a dedicated account with narrow permissions applies the same least-privilege controls for agents that security teams demand of any software holding its own logins.
On training data, Manus’s help center says it uses only aggregated or de-identified data to improve its services. The privacy policy has the full terms, so read it before uploading anything sensitive.
Who should use Manus?
Manus suits people who want a first draft of a deliverable and can verify it. A hypothetical example: ask it to pull published pricing from five named competitors into a spreadsheet with source links. The scope is clear, and every cell can be verified.
It fits poorly for:
- teams that can’t let a third-party agent touch sensitive systems
- anyone who needs a fixed monthly bill
- people who expect production-ready apps without testing
- people who mostly want conversational answers rather than finished artifacts
FAQs
Q. Is Manus free?
Partly. The free plan refreshes to 300 credits daily on the lighter Manus 2.0 Lite agent, which covers simple tests. New accounts also get a one-time 1,000-credit bonus, though the amount can vary by country. Paid plans start at $20 a month.
Q. Is Manus better than ChatGPT or Claude?
No independent head-to-head benchmark that this review could find settles it. Manus specializes in agent workflows that end in a built artifact, so the job should decide. OpenAI introduced its always-on dots agents on September 29, 2026, a day after Manus 2.0, which puts always-on delegation inside ChatGPT as well.
Q. What model does Manus use?
Manus offers Lite, 2.0, and 2.0 Max agents, plus Flex, which runs on your own model API key. It doesn’t publish which model powers which agent. An MIT AI Agent Index entry points to external models from providers including OpenAI, Anthropic, and Google.
Q. Is Manus owned by Meta?
No. Beijing ordered the acquisition unwound, and Manus has operated independently since September 1, 2026. Its October funding round was led by Boyu Capital and IDG Capital.
Related: 11 Best Agentic AI Frameworks in 2026: A Complete Decision Guide
| Disclaimer: This article is for informational purposes only and isn’t affiliated with Manus AI. Features, pricing, credits, and security details can change, so confirm anything critical through official Manus sources before deciding. |
