In March 2025, a short demo video started making the rounds on X. It showed an AI that could browse apartment listings, compare prices, and hand over a tidy spreadsheet without being told how. The agent’s name was Manus AI, and within 24 hours, everyone was asking for an invite code.
Most AI demos are smoke and mirrors. This one made people genuinely pause. Manus didn’t just generate text. It planned a task, broke it into steps, opened a browser, and got to work. It even paused to reason when it hit a dead end. The name is Latin for “hand”, and the team behind it built exactly that: an AI that does things instead of just talking about them.
What Makes Manus AI Different?
Most of the AI products you’ve used so far are chatbots. You ask, they answer. Even the advanced ones are essentially auto-complete with a pleasant tone. Manus sits on a different level. It’s an autonomous agent, which means it can complete multi-step tasks with minimal guidance.
- It plans tasks in steps rather than answering one-off questions.
- It interacts with live websites and software.
- It writes and runs code inside a cloud sandbox.
- It works asynchronously and sends updates when finished.
Those four capabilities give Manus a practical edge over typical chatbots. But the details are what make it interesting.
From Conversation to Execution
Give Manus a request like “find top-rated project management tools and compare their pricing”, and it will handle the legwork. It navigates websites, reads content, compares notes, and returns a detailed report. It’s not perfect. It can get stuck, and it sometimes moves at a speed that feels like watching someone type with one finger. But it learns. You can tell it “ignore the free trials” and it will re-run the search with that constraint in mind.
The bigger leap is that Manus operates in a cloud environment where it can write and run code, interact with web applications, and adjust its approach based on what it sees. That’s a far cry from a chatbot that simply predicts the next word.
The Benchmarks That Got Everyone Talking
The company claimed Manus outperformed OpenAI’s Deep Research on the GAIA benchmark, a test designed to measure AI agents on real-world tasks. Independent verification was thin when the hype peaked, but the numbers were enough to turn heads.
Manus also works asynchronously. You can close your laptop, and it keeps working in the cloud. When it’s done, you get a notification. In a world of synchronous back-and-forth chats, that feels genuinely useful.
Why Everyone Was Chasing an Invite Code
Manus launched with limited access. No public pricing, no API, just a waitlist and a sparse stream of invite codes. The scarcity created an immediate grey market. “Looking for a Manus AI invite” became a daily refrain on X and LinkedIn. Some invites reportedly sold for triple-digit amounts on resale platforms.
That kind of hype is typical for AI launches these days, but Manus had one notable distinction. It originated in China, not a Western lab. That gave it an air of unpredictability. The product emerged on Chinese platforms first, then spilled into the global conversation. It felt less like a polished marketing rollout and more like a spontaneous breakthrough.
The Open-Source Counterattack: OpenManus
Within days, developers released OpenManus, a stripped-down, fully open-source version of the same concept. It doesn’t have the polish or the scaling of Manus, but it proved a point: the architecture behind these agents is not magic. Anyone with a solid coding stack can build a rough equivalent. If you want to see how the project works, read our piece on OpenManus, the open-source AI agent that puts you in control.
That open-source push is part of a larger pattern. The race for capable AI agents is happening in the open, not just inside corporate labs. Communities are moving quickly, and open models are often closing the gap in a matter of days, not months.
A Geopolitical Flashpoint
You can’t discuss Manus AI without touching on geopolitics. The startup behind it, Monica, is based in China, and that has shaped its trajectory in ways that go beyond code. Reports emerged that Meta had offered to acquire the company, but Chinese regulators stopped the deal. The result was a clear message: China intends to keep its most advanced AI technology at home. For a detailed account, read our analysis of how China killed Meta’s acquisition of Manus.
This is not an isolated move. China has been quietly implementing policies to retain its top AI researchers, and projects like Manus are evidence that the strategy is working. China is increasingly keeping its best AI talent to itself, and that includes the researchers who built Manus.
Where AI Agents Are Headed Next
The Manus moment was a wake-up call. It showed that AI agents can move from research prototype to mainstream conversation overnight. It also exposed the real bottlenecks: coordination, memory, and trust.
Startups are already building on this idea in more commercial directions. Runable, which raised $21 million to bet AI agents can go from building businesses to growing them, illustrates the shift. That move from task-based execution to ongoing business ownership is where the next wave of agent products is headed.
Search is another domain that’s changing fast. Agents like Manus force you to rethink how information is gathered. Instead of scanning a list of links, you ask an agent to compare, decide, and return one polished answer. Shopify has already reported that AI-powered search drives more traffic and sales than traditional keyword search, which suggests that agent-led discovery is not a distant dream.
Manus AI, wherever it goes from here, cracked open a door that can’t be closed. We are no longer just talking to AI. We’re handing it a list of chores and trusting it to complete them. And once that habit forms, it’s hard to break.