Voltar às notícias
ProdutoInstruções AI Understanding

Manus lança plataforma de agente de IA com novas ferramentas e acesso antecipado

A startup chinesa de IA Manus revelou o Manus 2.0 e o aplicativo Cue, uma plataforma de agente de IA multifuncional que promete menor uso de tokens, conclusão mais rápida de tarefas e recursos integrados de agente pessoal.

4 min readRead the linked source
Source-page capture accompanying Manus launches AI agent platform with new tools and early access
Referência de fonteFonte registrada
Editora
uk.investing.com
Link da fonte
uk.investing.comhttps://uk.investing.com/news/stock-market-news/manus-launches-ai-agent-platform-as-startup-expands-tools-93CH-4886183
Tipo de fonte
Fonte vinculada — o status da fonte primária não foi estabelecido.
ContextoEntenda isso em 60 segundos

Comece aqui

Termos-chave

Agente de IA
Um sistema de software que pode observar, raciocinar e realizar ações para atingir um objetivo, geralmente usando ferramentas e memória.
Referência
Um teste padronizado ou conjunto de dados usado para medir e comparar o desempenho do modelo.
Calcular
Os recursos de processamento necessários para treinar e executar modelos, geralmente medidos em FLOPS ou horas de GPU.
Teste você mesmoQuestionário sobre agentes de IA

O que aconteceu

Manus announced the release of Manus 2.0, an upgraded AI‑agent system called Cascade that uses 23.2 % fewer tokens, finishes tasks 28.2 % faster and costs 32 % less in a single test configuration. The rollout adds a Cloud Computer environment for persistent hosting, Automations that trigger tasks from services like email, Slack or Notion, and a revamped desktop client called Manus Studio with Video Editor and Game Dev workspaces. In parallel, the company launched Cue, a standalone app that assigns each agent its own email address, phone number, wallet and virtual computer, enabling agents to send messages, make budget‑constrained payments and operate independently. Cue also supports grouping agents to collaborate on shared goals. Both Manus 2.0 and Cue are available now on web, desktop and mobile; iOS support for Cue will follow App Store approval, and early‑access users can join for free using the invite code MEETCUE on a first‑come, first‑served basis.

Manus, a China‑based artificial‑intelligence startup, released Manus 2.0 on 28 September 2026. The core of the update is the Cascade agent system, which the company says reduces token usage by 23.2 % and cuts task execution time by 28.2 % while lowering operating costs by 32 % in a single configuration.

The platform now includes Cloud Computer, a dedicated hosting environment for projects that need persistent resources, and Automations that let users trigger AI‑driven actions from external services such as email, Slack or Notion. The desktop client, renamed Manus Studio, adds two new workspaces: Video Editor for short product ads and tutorials, and Game Dev for building and publishing multiplayer games with built‑in hosting via Cloud Computer.

Alongside Manus 2.0, the company introduced Cue, a standalone application that assigns each its own email address, phone number, cryptocurrency wallet and virtual computer. Cue enables agents to send messages, make budget‑limited payments and operate independently, and it supports grouping agents to collaborate on shared objectives. Cue is currently available on web, desktop and mobile platforms, with iOS support pending App Store review. Early access is free using the invite code MEETCUE on a first‑come, first‑served basis.

Detalhes da fonte: uk.investing.com ↗

Por que isso importa

The launch marks a notable step toward more autonomous, personal AI agents that can act on behalf of users across communication, finance and productivity workflows. By reducing token consumption and execution time, Manus 2.0 lowers the cost barrier for developers and enterprises seeking to embed AI agents in products, potentially accelerating adoption in sectors such as marketing, education and gaming. Cue’s provision of dedicated contact channels and wallet access blurs the line between software bots and independent digital actors, raising both opportunities for streamlined automation and concerns about misuse, fraud or regulatory oversight. The early‑access model and free tier suggest the company aims to build a user base quickly, but the lack of disclosed pricing beyond the invite phase leaves the long‑term commercial viability uncertain. Independent verification of the claimed efficiency gains is pending, making real‑world performance and scalability key questions for adopters.

The reduction in token consumption and execution time directly translates to lower costs, making AI‑agent deployment more affordable for startups and larger enterprises alike. This could spur a wave of new applications that embed autonomous agents in everyday workflows, from marketing content creation to real‑time game moderation.

Cue’s provision of personal communication channels and a wallet for each agent creates a functional digital persona capable of interacting with external services without human mediation. While this expands the utility of AI agents, it also introduces regulatory and security challenges, as autonomous agents could be exploited for phishing, fraud or other malicious activities if safeguards are insufficient.

The free early‑access model suggests Manus is prioritizing rapid user acquisition and community feedback. However, the absence of disclosed pricing beyond the invite phase leaves uncertainty about the platform’s long‑term revenue model and whether it can sustain the infrastructure required for persistent hosting and real‑time agent interactions.

Interactive Mechanism

Mecanismo interativo: como realmente funciona

Explore a tecnologia subjacente a este desenvolvimento de forma interativa.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:🛡️ Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Verificação de conceito interativo+10 Points
AI Agents Quiz

An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?

O que assistir a seguir

Key indicators to monitor include the conversion rate from free early‑access users to paid subscribers, any announced pricing tiers for enterprise usage, and how regulators respond to agents that can send messages and make payments autonomously. Additional scrutiny will focus on security audits of the wallet and communication features, the rollout of iOS support, and whether third‑party developers integrate Manus 2.0 into larger platforms or create novel use cases that could amplify both productivity gains and potential abuse vectors.

Adoption metrics: conversion from free early‑access users to paying customers, and the scale of enterprise contracts that may follow.

Pricing announcements: any future disclosure of tiered pricing for cloud hosting, automation triggers, or premium agent capabilities.

Regulatory response: potential guidance or restrictions from financial regulators concerning agents that can make payments, as well as data‑privacy considerations for agents handling personal communication channels.

Security audits: independent assessments of the wallet integration, email/phone provisioning, and overall sandboxing of agents to prevent abuse.

Platform expansion: rollout of iOS support for Cue, integration with additional third‑party services, and the emergence of developer ecosystems building on Manus 2.0.

Guias e questionários relacionados

Agentes de IAÉtica da IAFuturo da IAModelos de IA explicadosTeste o que você sabe – experimente um teste gratuito de IAProcure um termo de IA em nosso glossárioSiga o rastreador de lançamento de modelo de IA
Achou isso útil?