# Tencent Tests OS-Level Personal AI Assistant Tencent Marvis

- Category: Flash (快讯)
- Published: 2026-05-19T11:56:00.000Z
- Source date: 2026-05-19
- HTML: https://www.88z.ai/en/news/flash-20260519-46a36d284c8e10ee
- Markdown: https://www.88z.ai/en/news/flash-20260519-46a36d284c8e10ee.md
- Other language: https://www.88z.ai/news/flash-20260519-46a36d284c8e10ee.md

Tencent's App Bao team has launched internal testing for Tencent Marvis, an OS-level AI assistant. It supports the local deployment of six collaborating agents and uses natural language commands for file management, system settings, and cross-modal retrieval, balancing cloud efficiency with local privacy.

## At a glance

- Supports local deployment of six collaborating agents to handle various tasks via natural language commands
- Operates in a dual cloud-local mode, requiring L2-level hard inquiry confirmation for sensitive operations

## The update

Tencent's App Bao team is conducting internal tests on Tencent Marvis, an operating system-level personal AI assistant. The product allows for the local deployment of six AI agents on user computers, enabling them to collaborate on complex tasks. Users can manage files, adjust system settings, perform cross-modal content retrieval, and automate multi-task execution simply by using natural language commands. In terms of operation, Tencent Marvis adopts a dual mode that combines cloud efficiency with local privacy protection. To ensure data security, the system triggers a mandatory L2-level hard inquiry confirmation process for sensitive operations. This ensures that critical instructions are executed only with user knowledge and authorization, thereby strictly maintaining privacy boundaries while improving work efficiency.

## Sources

- [腾讯科技](https://mp.weixin.qq.com/s/UEKg2imwHA8hzV1kBR6e1g) (2026-05-19)

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.