which only announced its resumption of independent operations on September 1 Manus, and returned less than a month later with a major version update. This time, Manus 2.0 it launched three offerings at once: an upgraded desktop app Manus Studio, a brand-new personal agent application Cue, and an in-house next-generation Agent framework CascadeThe team that once brought the concept of “general-purpose Agent” into the spotlight has now equipped AI with a phone number, a wallet, and a computer—and enabled it to form groups and collaborate on tasks.
What does Manus 2.0 deliver?
In short, Manus 2.0 represents an upgrade from “point tools” to a “complete product ecosystem.” It integrates office work, content creation, entertainment, and long-term tasks onto a single infrastructure, supporting different tasks via two distinct interaction modes:
- Cue: the brand-new personal agent application, equipping Agents with email, phone number, wallet, and computer—plus multi-assistant group collaboration.
- Manus Studio: the upgraded version of the original desktop application, now adding video editing and game development capabilities beyond documents, spreadsheets, PDFs, and presentations.
- Cascade: a next-generation in-house Agent framework; official benchmarks show a 23.2% reduction in token consumption, a 28.2% decrease in task completion time, and a 32% drop in runtime cost.
This launch primarily targets overseas users; the company also revealed that its domestic-market product is under preparation and its local team is being assembled. Notably, the release date falls exactly 10 days after Meta’s personal Agent Muse topped the U.S. App Store Free Charts—turning the former acquirer and acquired into direct competitors in the Agent race.
Cue: Equip Agents with phone numbers, wallets, and group chats
Cue is the most eye-catching release this time. It is a standalone app sharing infrastructure with Manus, available on both mobile and desktop. Each Agent in Cue hasan independent identity—including an email address, phone number, wallet, and its own computer.
With these resources, Agents gain the fundamental capability to “get things done online” like human assistants: they can send messages externally, complete payments within user-defined budgets, execute tasks on their own machines, and even answer calls on behalf of users and generate call summaries. Think about how humans operate online—email and phone numbers let others find and contact you; payment tools enable transactions; computers handle various tasks. Cue equips Agents with this entire set of capabilities.
And you can have more than one assistant. Users can assign distinct responsibilities to different Agents and bring them together into the same group chat for collaboration. The official example is organizing a New York launch event: one Agent handles venue scouting, another compiles candidate lists, and a third drafts presentation slides—passing work along toward a shared goal, while the user sets direction and makes final decisions. Launch materials also demonstrate real-world extensions, such as scanning a QR code at a restaurant to let an Agent place orders or join queues.
Cue is currently inearly accessand requires an invitation code to use.
Manus Studio: Your video is edited—now you can tweak it yourself
Upgrading the desktop version to Studio unlocks the most significant expansion in creative capabilities. For video, Studio provides a hands-ontimelineeditor: clips, images, text, animations, and audio are placed separately, allowing users to swap assets, adjust positioning, export directly after edits, or hand modified versions back to Manus for further processing. This design embeds “human intervention” into the content creation process—the closer the output is to final form, the more critical it becomes to have a direct interface for those small, last-minute adjustments.
The video editor targets 30- to 60-second content such as product ads, tutorials, vlogs, and data-driven motion graphics; Alchemy mode simultaneously leverages video generation and code generation for ideation, animation, and editing. Game development follows a similar logic: Game Dev offers playable templates powered by image, video, and programming models—users can watch games run in real time while editing code and tuning characters or scenes, then publish the result as a web page friends can open and play instantly; for multiplayer, a continuously running server is required, and Manus integrates cloud PC procurement and multiplayer configuration directly into the workflow.
Personal Agent race: The wind is blowing again
By the time Manus 2.0 launches, the personal Agent space is already crowded with new entrants. Meta launched its personal Agent Muse on September 8, emphasizing memory of users and persistent goal pursuit—soon topping the U.S. App Store’s free chart; formerbean bag PC division head Qi Junyuan entered the market with Today AI, similarly highlighting long-term memory and proactive action; OpenAI is also reportedly developing a persistent Agent. The distinction between personal and enterprise Agents goes beyond context length—enterprise deployments prioritize permission management and data isolation, whereas personal Agents emphasize sustained understanding of an individual’s preferences, schedules, and goals.
For everyday users, the immediate benefit of this competition is clear: an AI assistant that stays with you long-term, remembers you, and runs errands on your behalf is shifting from concept to usable product.
Frequently Asked Questions (FAQ)
Is Manus 2.0 free?
Manus has not yet announced the full pricing structure for Manus 2.0. Cue remains in early access and requires an invitation code to use; Manus Studio retains its prior product format, and specific pricing plans should be confirmed via official announcements.
What is the difference between Cue and Manus?
Cue is a new personal agent application that equips agents with phone numbers, digital wallets, computers, and group chat collaboration—positioning itself as a long-term personal assistant; Manus Studio is a desktop workspace focused on productivity and creative tasks (e.g., video editing, game development). Both share the same underlying infrastructure.
Does Manus 2.0 support Chinese? Is it available in China?
This release primarily targets overseas users, but the official team has confirmed that a China-specific product is under preparation and its team is being assembled. Manus previously supported Chinese-language interaction, and this capability is expected to remain in the domestic version upon launch.
What can Manus Studio do?
Beyond standard office capabilities—such as handling documents, spreadsheets, PDFs, and presentations—Manus Studio now adds video editing (30–60 second product ads, tutorials, vlogs) and game development (playable templates, web deployment), with real-time user intervention and editing at any stage.
Manus and GPT-6.1 Sol,Claude What is the relationship between such models?
Manus operates at the product layer (an agent application) and calls upon various large models at the infrastructure level; GPT-6.1 Sol,Claude and others reside at the model layer. Manus’s value lies in orchestrating model capabilities into end-to-end, task-executing workflows. For details on underlying models, see our GPT-6.1 Sol benchmark and Claude Sonnet 5.5 Review.
Summarize
The significance of Manus 2.0 lies in completing the evolution of a “task-performing AI” into a full-fledged agent—one with identity, contactability, payment capability, and collaborative functionality. Cue’s group chat collaboration and Studio’s real-time intervention both point in the same direction: AI is no longer just a question-answering tool, but an active partner that executes tasks on your behalf—and one you can seamlessly take over and modify at any time. Whether this system delivers truly smooth, everyday usability remains to be seen in subsequent deployments.
Want to Discover More Useful AI Tools? Explore Our AI Model Library and Tool Comparison Engineor continue reading:Cohere Command A+ Review · Xiaomi MiMo Code Review · Kimi K2.7 Code Evaluation · Grok 4.7 review · DeepSeek V4 Pro benchmark.