The AI agent everyone was talking about in early 2025 — what it is, what it can do, and why it's caught in a battle between Meta and the Chinese government.
The AI agent everyone was talking about, that Meta is trying to get their hands on
In March 2025, every AI-enthusiast was going on about one thing.
Twitter, LinkedIn, every smug-tech-part of the internet, people were going on about this new tool that had swept in from China and was going to change the face of Agentic AI as we know it.
Manus.
Now just to be completely honest with you, I personally don’t remember any of this hype. I wasn’t interested in AI back then and my social media was mostly funny animals and toddlers falling over.
Manus promised a whole new way of interacting with AI and slick teaser videos showed incredible use-cases which made Silicon Valley (rightly) sit up straight.
But what was about Manus that got everyone so excited? What are they up to now? And did they really get acquired by Meta??
Lots of questions, time for some answers!
What is Manus?
Manus is a general-purpose AI agent, built by a Chinese startup called Monica (sometimes also referred to as Monica.im).
It launched in early March 2025 and went viral almost instantly, largely because it was doing something that most AI tools weren’t doing at the time.
Actually doing things.
So opening files, creating documents, linking to tools…everything without requiring any further input from a human.
Now, writing this in April 2026, this might not sound like such a “wow” thing. After all, nowadays there are many AI Agent products on the market for users to choose from, or use n8n to build their own.
But back in early 2025, most AI tools were purely conversational LLM models.
You spoke to them and they spoke back. (Claude Code wasn’t in fact released until January 2026)
So there was a big race to be the first to create truly autonomous agentic tools. Tools that could perform real-life tasks from start to finish.
PS: If you aren’t familiar with AI Agents, you can take a look at my article on AI Agents.
Manus swept on to the scene with a promise of an AI Agent that could do real tasks like: screening resumes, creating trip itineraries and analysing stocks.
In true Berlin fashion, they also hyped it up even MORE by making the initial access “invite only”.
Alongside the excitement was a good deal of scepticism and caution.
Manus was developed by a Chinese company and there was rightly a great deal of talk about safety and data privacy concerns.
Manus can perform tasks by gaining access to a range of systems. Since this is running on the Cloud rather than locally on your computer, there is a greater risk of security breaches and privacy issues.
What Can Manus Actually Do?
According to the hundreds of hyped up YouTubers I scrolled through, Manus can do a lot and is maybe the best thing in the world since sliced bread.
(A fair few of these YouTubers were sponsored by Manus though, so you really have to filter to get to the honest stuff)
Of course it has all of the chatbot style features that we are used to by now from ChatGPT or Claude — you can ask it questions and it can answer them for you.
That’s straightforward enough.
But where Manus really comes into its own is performing full tasks from start to finish. Some use-cases that stand out are:
- Full website creation — describe the website you want and Manus builds it end to end. It writes the code, structures the pages, and delivers a working site without you touching a single line of code.
- App creation — give Manus a brief for an app and it plans the architecture, writes the code, and produces a working prototype.
- Presentation slide creation — tell Manus the topic and the audience and it researches the content, structures the narrative, and produces a finished slide deck ready to present.
- End-to-end market research reports — point Manus at a research question like “analyse my three main competitors” and it searches the web, visits websites, pulls pricing and feature information, synthesises everything, and delivers a formatted report with charts and insights.
- Candidate screening and recruitment — feed Manus a job description and a folder of CVs and it reads through all of them, scores candidates against your criteria, and produces a ranked shortlist with notes.
For me, while I was testing Manus, I asked it to do a full SEO and general review of this learning platform.
It ran for quite a while but with zero input from me, the output was a finished PDF file with a full SEO analysis of my site, with pretty useful suggestions of how I could improve it.
I was fairly impressed!
How to get started with Manus
You can get started right away with Manus by going to their site at https://manus.im/ and creating a free account.
As I write this, new users get 1000 tokens free to play around with, which is nice, and they do technically offer a free version (but with only a small number of tokens each day).
I would say, create an account, play with the 1000 free tokens and you will get a fairly good idea of what Manus can do.
The Pros
1. There is a cool transparency feature
One of the things that makes Manus really stand out is its transparency feature.
As it works through your task, you get a live log of every step it’s taking.
You can watch it actually open a browser tab, run a search, read a page, write a bit of code, commit it to git…
This is quite a cool feature because most AI tools really work like a black box. You never get to really see what they are doing and this (at least for me) can lead to distrust of how they really got to the end result.
With Manus you can scroll back through and see exactly which steps it took and why.
Quite a unique and interesting feature!
2. It handles complex, multi-step tasks without hand-holding
For research-heavy work especially, Manus is impressively capable.
Give it a chunky task and it will just crack on with it, rather than stopping every two minutes to ask a clarifying question.
Personally this bothers me a fair bit with Claude… it asks me for my opinion wayyy too much. Just do it already!!
This is really cool for having tasks run in the background or in parallel, without having to keep checking or reviewing the next steps.
3. No coding or setup required.
The lack of coding required is a big plus for a lot of users. If you are capable of typing a task in plain English, you can still do an awful lot with Manus.
Manus’ biggest competitor n8n is very cool and effective but it still requires a fair amount of setup and technical know-how.
Manus really is a type-and-watch-me-roll tool.
The Cons
1. Privacy, privacy, privacy!
Manus is built by a Chinese company, and your tasks, files, and data are being processed on their servers.
If you are working with anything sensitive — personal data, confidential work documents, anything you wouldn’t want sitting on a foreign server — you should think carefully before using it.
2. Security
Because Manus operates inside its own computer environment and can browse the web, run code, and interact with files, the question of what you give it access to matters.
It is sandboxed, which helps, but no system is completely immune.
This means that for your average Joe, maybe there’s no problem, but for a corporation who handles millions of lines of sensitive user data? Hmm… I would be slightly more careful here.
(PS: As I write this Manus is in the middle of a fight-for-the-soul battle between Meta — who want to acquire it — and the Chinese government who don’t want the tech falling out of their control. Spicy stuff)
3. Cost
Manus is not free and according to the reviews I have been extensively reading…it’s EXPENSIVE.
People seem to complain that Manus eats up tokens very quickly and this can lead to the costs quickly racking up.
For my small task of “SEO analysis of a website”, it used nearly 200 tokens of my free 1000 token limit.
I could imagine this getting very hungry, very fast when it comes to really complex multi-agent tasks.
My final thoughts on my experiences with Manus. I think that this is definitely a tool to keep an eye on in the future.
As I write this, Manus AI are caught in a fierce battle between the Chinese Government and Meta, the latter who have been trying to acquire it for over a year and the former who don’t want their technology being handed over to a US company.
I am imagining what an AI Agentic tool could be capable of, if given free access and integration to all of Meta’s tools and platforms.
Instagram content could be automatically generated and released…
WhatsApp replies for business could be completely run using AI Chatbots….
If this acquisition does go ahead, I think it could spell big changes for the world of content generation, using AI Agents.
Summary
- Manus is a general-purpose AI agent — not a chatbot that answers questions, but an AI that autonomously plans and completes multi-step tasks on your behalf.
- An AI agent acts, it doesn’t just answer — give it a goal and it figures out the steps, carries them out, and comes back with a finished result.
- Manus works inside its own computer environment — it can browse the web, write and run code, manage files, and produce finished outputs without you directing every move.
- Manus AI was created by a Chinese company and is embroiled in a legal battle with Meta over acquisition. The outcome of this legal fight could have serious implications for the security/privacy and overall future of the tool.