Manus AI

Manus is an autonomous AI agent designed to do more than answer questions. You give it a goal, and it can plan the work, use tools, browse the web, operate a computer environment, create files, build digital products, and return a finished result.

That positioning has become even clearer with Manus 2.0, introduced on September 28, 2026. The update adds a new Cascade agent harness, Cloud Computer, Automations, Manus Studio, and Cue. Manus says the new architecture is designed to connect research, creation, automation, and ongoing projects rather than treating every request as an isolated chat. Manus

There is also an important company-history detail for anyone researching Manus today. After a proposed acquisition by Meta was unwound, Manus formally resumed independent operations in September 2026. The company now describes itself as an independent agent lab. Manus

Manus Is Built Around Delegating Work, Not Just Generating Text

The easiest way to understand Manus is to compare it with a normal chatbot.

A chatbot usually waits for a question and produces a response. An autonomous agent can take a broader objective and decide which steps are needed to reach the requested result.

That distinction is central to Manus.

For example, a request can involve researching information, organizing the findings, creating a presentation, generating supporting assets, or building a website. Manus can work through multiple stages instead of stopping after the first written answer.

The product’s current website highlights several major workflows, including website creation, slide generation, design, image generation, music generation, browser operation, Wide Research, Mail Manus, and Slack integration. Manus

This makes Manus less like a single AI writing tool and more like a workspace where an agent can use different capabilities depending on the task.

That approach is particularly useful when the desired result is a deliverable rather than a paragraph of text.

Manus 2.0 Changes the Shape of the Product

The biggest current development is Manus 2.0.

The company says the release introduces Cascade, its latest in-house agent harness. According to Manus, Cascade dynamically brings specialized capabilities into a project when they are needed. The company reports that, in one tested configuration, Cascade used 23.2% fewer tokens, completed tasks 28.2% faster, and cost 32% less to run than its previous system. Those figures are company-reported results from a particular configuration, not an independent benchmark. Manus

More interesting than the efficiency numbers is the broader product architecture.

Manus 2.0 adds Cloud Computer, which gives projects a dedicated computing environment. This matters for tasks that need to continue running after the user’s local computer is closed. Manus specifically describes use cases such as persistent servers and automations that need an environment available around the clock. Manus

Then there are Automations. Earlier scheduled tasks could start work at a particular time. The new system can also react to events from connected services. Manus gives examples such as new emails, advertising changes, calendar events, Slack messages, or Notion updates triggering an agent workflow. Manus

That moves Manus closer to an automation platform rather than a simple on-demand assistant.

Research, Browsing and Building All Sit in One Workflow

Manus has long been positioned around autonomous research, but the current product goes further.

Its browser capabilities allow the agent to interact with websites rather than merely summarize search results. Its Wide Research feature is designed for larger research tasks that can be divided across multiple directions.

This matters when the job involves collecting information from several sources and turning it into something usable.

A typical workflow might start with a broad research request, move into web investigation, organize findings, create a report, and then turn that research into slides or another deliverable.

The company has also expanded its connector ecosystem. Current product material lists integrations and connectors involving services such as Google Drive, Gmail, Google Calendar, Slack, Zoom, Shopify, Notion, Airtable, HubSpot, Canva, Similarweb, and others. The exact availability of connectors can depend on the workspace and product configuration. Manus

This is one of Manus’s more useful characteristics: the agent does not have to remain isolated from the tools where work already happens.

Websites and Slides Make Manus More Than a Research Agent

One reason Manus attracts attention is that its output is often tangible.

The platform can create websites and web applications, generate presentations, produce images, and work with data. The official product navigation currently exposes dedicated AI design and AI slides products alongside the main agent, while recent updates have added PowerPoint support, design views, custom domains, app sharing, and publishing-related features. Manus

For someone who wants to turn an idea into a prototype, this can reduce the number of separate tools involved.

The same applies to presentations. Instead of asking an AI to outline ten slides and then manually designing them, Manus can work toward a finished deck. Its product updates also show a continued focus on editable presentations and professional creation workflows. Manus

There is still an important distinction, though. Generating a working first version is not the same as producing a final client-ready artifact every time. Autonomous systems can make incorrect assumptions, miss requirements, or produce work that needs human review.

That is a normal limitation of delegating open-ended work to an agent.

The Developer Side Includes an API

Manus is not limited to its consumer-facing interface.

The current Manus API v2 allows developers to programmatically create and manage agent tasks. Official documentation describes task creation, follow-up messages, files, projects, webhooks, skills, and custom agents. API v1 is deprecated, with v2 now designated as the current version. Manus API

This opens up a different use case.

Instead of asking a person to manually start Manus every time, a developer can integrate agent execution into an existing application or workflow.

The API supports files as task attachments and webhooks for receiving notifications when tasks complete or require input. Manus also provides an official API skill that can be installed into compatible coding agents to keep integration guidance aligned with the live API documentation. Manus API

For technical teams, this makes Manus more interesting than an ordinary AI productivity subscription.

The trade-off is that API integration requires more technical setup and careful management of permissions, credentials, task execution, and usage.

Pricing Is Credit-Based, So Usage Matters

Manus uses a credit-based model rather than simply offering unlimited autonomous work.

The official Help Center currently lists a Free plan at $0 per month, with Chat Mode and Manus 1.6 Lite in Agent Mode, 300 daily refresh credits, one concurrent task, and two scheduled tasks. Manus Help Center

The official Help Center also lists Pro options starting at $20 per month for 4,000 monthly credits and $40 per month for 8,000 monthly credits, with annual billing offering a stated 17% discount. These Pro options include access to Manus 1.6, 1.6 Max, and 1.6 Lite in Agent Mode, along with advanced research, website deployment, slide generation, Wide Research, and higher task limits. Manus Help Center

For teams, the Help Center lists a Team plan starting at $20 per seat per month, also with a stated annual-billing discount. Team features include SSO, usage analytics, internal access control, shared slide templates, and a data-training opt-out. Manus Help Center

There is a complication: Manus’s public pricing page is currently heavily JavaScript-driven and does not expose the detailed tier table in the text available to search engines. The official Help Center is therefore the clearer source for the verified plan information above. Manus

The practical lesson is simple: don’t evaluate Manus only by its monthly subscription price. Evaluate how quickly your expected tasks consume credits.

Where Manus Fits Best

Manus makes the most sense when the task has multiple stages and the user wants the system to perform those stages rather than simply provide instructions.

Research projects are an obvious example. So are first-draft websites, presentations, data collection, competitive research, workflow automation, and prototype development.

It is also becoming more useful for recurring work. Automations can react to events, while Cloud Computer provides a persistent environment for projects that need to remain available. Manus

At the same time, Manus is not automatically the right tool for every AI task.

If you only need a short answer, a conventional chatbot can be simpler. If you need highly deterministic workflow automation, a traditional automation platform can provide more explicit control. If your work is exclusively software engineering, a dedicated coding agent may offer a more focused environment.

The value of Manus comes from combining many of those activities inside one agent-oriented workspace.

And Manus 2.0 makes that direction even clearer. It is no longer just about asking an AI to perform a complicated task. The product is moving toward projects that can continue working, interact with connected services, create deliverables, and respond to events.

That is the part worth watching.