The latest AI news and updates of Manus — an autonomous general AI agent that executes tasks and automates workflows. Covering Manus's latest product updates, company news, and launches from the past 90 days.
ManusSep 9Manus has integrated OpenAI's GPT-Image-2.5, reporting that OpenAI's Flare variant generates images two to four times faster than GPT-Image-2. The update also introduces improved transparent-background generation, enhancing the agent's ability to create brand assets, presentations, and web content.
ManusAug 11Manus is resuming independent operations following its acquisition by Meta. To comply with regulatory requirements, some users must back up their data before 7:59 a.m. SGT on August 23, 2026, and restore it starting at 8:00 a.m. SGT on August 25, 2026. Affected accounts will not be charged during this transition and will receive welcome-back bonuses.
ManusJul 22Manus launched Plan Mode, a manual review step that pauses execution to generate a structured plan. The agent researches feasibility and surfaces edge cases before building assets. The plan requires manual confirmation before the agent begins work, preventing unintended changes. This feature is available now on web and mobile platforms for all users.
ManusJul 20Manus updated its Google Workspace connector to support multi-account functionality. The integration allows the AI agent to link up to 10 different Google accounts simultaneously. This update enables the agent to manage workflows across multiple workspaces without requiring manual account switching.
ManusJul 17Ascendea deployed Manus as an AI chief of staff, named James, to coordinate its operations. This three-layer system handles 28 to 33 live workflows for 112 clients, resulting in a 90x increase in workforce output. The agent manages complex strategy and technical architecture, while local agents execute repetitive tasks, allowing the 10-person team to focus on final judgment.
ManusJul 16Manus launched a /typst-pdf-maker skill that converts natural language prompts into professionally typeset PDFs. The agent automatically handles document structure and layout for reports, research papers, resumes, and books. This capability generates polished, print-ready documents directly from a single prompt without manual formatting.
ManusJul 14Manus now generates native .pptx files directly, bypassing web-based conversion to preserve formatting and structure. The update introduces editable charts that redraw in real-time when data values change, alongside toggleable chart elements like labels and gridlines. PowerPoint mode supports inputs from spreadsheets, research, or documents and is available now in beta for 1.6 and Max models.
ManusJul 13Manus launched auto-publish, a toggle that automatically deploys successful website builds to a live URL. The feature remains off by default and integrates with message queuing, allowing multiple prompts to stack while the agent builds and ships each version sequentially. Auto-publish is available now across web, iOS, and Android platforms.
ManusJul 10Manus added a /game-dev skill that builds and deploys games from natural language descriptions. The agent handles the entire development process, generating working gameplay and hosting the project at a shareable link. This capability requires no prior game development experience and builds on open-source work by @alex_erm.
ManusJul 10Manus launched an ElevenLabs connector, enabling AI agents to integrate voice synthesis and cloning into autonomous workflows. The integration supports full-stack app development with multi-language narration. Alan, a Manus team member, demonstrated the capability by building a personalized song generator featuring a retro interface, database persistence, and voice cloning in a single prompt.
ManusJul 9Manus launched Branch, a feature that splits any conversation into a parallel session while keeping the original intact. Each branch inherits the full context, including instructions and files, to support multiple workstreams from a single research point. Branching is available now for all standard Manus chat sessions, with a breadcrumb trail for navigation between sessions.
ManusJul 8Manus upgraded its Google Drive connector to a full Google Workspace integration, now covering Docs, Sheets, Slides, and Forms. The update introduces granular access controls, allowing permissions to be set to Off, Read, or Full for each connected account. This provides precise control over how the AI agent interacts with files and data across the Google ecosystem.
ManusJul 1Manus launched a toolkit for small businesses, featuring connectors for Shopify, HubSpot, Yelp, Toast, and Zoom. The package includes 15 ready-to-run agentic workflows for website building, SEO optimization, and social content creation. Manus also introduced free webinars and a use case competition for businesses deploying these automated operational workflows.
ManusJun 25Manus launched a Semrush connector, allowing its AI agent to pull SEO data through natural language prompts. The integration covers keyword research, competitor gap analysis, and backlink profiles. Manus structures results into dashboards, charts, or slide decks and supports scheduling recurring reports for delivery to Slack or Google Docs.
ManusJun 24Manus launched Hosting Modes for its AI website builder, allowing projects to toggle between two server configurations. Autoscale scales from zero to five instances and costs nothing when idle, while Reserved maintains a persistent instance for real-time apps like streaming or bots. Users switch modes in project settings with one click, requiring no code changes or infrastructure management.
ManusJun 23Manus now suggests relevant web pages for project sources based on natural language descriptions. Instead of manually hunting for URLs, the agent identifies pages matching the specified needs. Once selected, these sources are integrated into the project context, allowing the agent to reference them directly during subsequent autonomous task execution.
ManusJun 16Manus now supports message queuing, allowing follow-up prompts to stack while the agent executes a current task. This feature eliminates the need to wait for task completion before capturing new instructions. Queued commands execute sequentially after the current task finishes, enabling batch processing for overnight builds or mid-build idea capture.
Frequently asked questions