Tencent’s personal AI assistant for local files, computer tasks, and mobile access
Marvis is Tencent’s personal AI assistant for completing tasks across a computer’s files, system settings, browser, and installed applications. It interprets natural-language requests, breaks work into steps, and returns results, alongside everyday question answering, search, summarization, and writing.
Windows and Mac computers serve as its main execution environment. Android and iOS apps provide a portable entry point for starting tasks, receiving results, and continuing work away from the desk. Marvis is designed for everyday computer assistance and information processing; complex workflows may still require specialist software or user involvement.
Find local documents and images using content rather than filenames. Search includes text within images, while AI document and image libraries organize material by attributes such as people, topics, locations, and holidays.
Read and analyze local documents and spreadsheets, refine writing, generate charts, and convert file formats to turn existing material into usable outputs.
Use natural language to inspect hardware and system information or change supported computer settings without navigating configuration menus.
Combine browser operations and installed applications into task workflows. Marvis decomposes requests and invokes the relevant agents, requesting confirmation for actions involving privacy, security, or payment.
Connect a phone to a computer to initiate work remotely, view the live execution screen, inspect results, and take control when needed.
Efficiency mode combines device and cloud processing, with complex understanding and planning handled in the cloud when needed. Local privacy mode uses an on-device model for files, image recognition and conversation; it also supports local-file tasks offline. Mode selection depends on computer hardware.
Bring scattered documents and spreadsheets together for contract information review, operational data analysis, or preparation of a deliverable.
Organize literature, distill books, and extract useful points from personal notes so existing materials can support study and writing.
During travel or meetings, request material from a connected computer and continue an interrupted task through the phone.
Use content-based organization to assemble collections such as a child’s growth album or locate images associated with a particular place or event.
Windows 10 or later on x64 hardware, a CPU with at least four cores, at least 8 GB of memory, and a solid-state drive are required.
Marvis requires an Apple Silicon Mac with an M1 chip or later and macOS 13 or later.
The mobile-only experience supports conversation, content organization, web search, and lightweight result generation. Connecting a Windows or Mac computer provides the broader local-file, system, browser, and application capabilities.
Yes. In the Privacy and Security settings, you can exclude folders from indexing. Document and image libraries initially show filenames and lists; recognizing file contents requires further authorization. Initial recognition can increase CPU usage and take time depending on file volume and hardware.
Local privacy mode can search and process local files, recognize image text and organize folders offline. Tasks that inherently need the internet, such as web search, webpage retrieval or current-price lookup, still require a connection. Efficiency mode uses cloud processing for complex understanding and planning.
Traffic data is for reference only.
| Jun | 720323 |
|---|---|
| Jul | 293463 |
| Aug | 236750 |
Traffic analytics are not available yet.
honeybear.ai
AI document assistant for searching, summarizing, and chatting with files and YouTube content
kaagaz.app
Kaagaz is a mobile-first suite for scanning documents, editing PDFs, backing up files, making posters, and creating resumes.
skilldocs.dev
SkillDocs is a collaborative markdown workspace for editing, commenting on, and sharing the documents that agents run on. It combines live multiplayer editing with an MCP-based workflow for pushing skills from local files, collecting feedback, and pulling changes back.
