ByteDance's AI product TRAE has released a personal assistant for developers named TRAE Bao. This feature became available on October 29th. TRAE announced the launch through its official channels and on the official Chinese community forum.
Unlike the previous format, where the assistant performed one task per session, the new tool is designed to track the developer's projects, tasks, and decisions over time.
TRAE explains the problem as follows: a significant portion of the time developers spend with AI-based coding tools is not spent writing code, but rather on project detailing, task distribution, and switching between work sessions. With every new task, they have to re-explain the technical stack, project background, and previously made decisions. Complex requirements must be manually broken down, tracked across sessions, and then compiled.
The assistant uses a two-layer architecture. The global assistant is responsible for coordination and storing memory across different projects. It can analyze past tasks, dialogues, and the developer's work results, extracting reusable solutions and constraints. When a developer starts new work and defines a goal, the global assistant automatically loads the relevant context by studying experience from earlier projects.
The project assistant, conversely, focuses on one project at a time. Working within the context of that specific project, it breaks down a complex need into a series of actionable tasks, runs subtasks in parallel, integrates them, and summarizes the results. According to TRAE's description, the developer's role in this process shifts towards receiving reports and making decisions.
In the desktop application, access to the global assistant is available through a dedicated entry point, while creating or opening a project activates the project-level assistant. The global assistant can be integrated with messengers such as Feishu and WeChat, allowing developers to assign tasks and receive progress updates without opening the client application.
There are certain limitations: the current version only supports local tasks. TRAE has stated that support for cloud tasks and mobile devices will appear in future updates. Users on the official forum note that the feature is being rolled out gradually, first appearing in the Chinese version, and that an automatic model selection is currently used without manual configuration capability.
This launch adds TRAE to the growing group of desktop agents that allow users to delegate work from phones or chat applications and maintain long-term memory between sessions.
