ByteDance Consolidates TRAE and Coze into Doubao, Readies 'Doubao Work' Enterprise Brand

ByteDance has initiated an internal organizational restructuring to consolidate its enterprise AI developer tools and agent platforms under the Doubao ecosystem. The company is merging the teams and technologies behind its AI programming suite TRAE and agent-building platform Coze into Doubao, preparing to launch a unified enterprise AI suite branded "Doubao Work." The restructuring concentrates disparate AI tools into a single corporate pillar to compete against domestic rivals, notably Tencen

2 min
ByteDance Consolidates TRAE and Coze into Doubao, Readies 'Doubao Work' Enterprise Brand

ByteDance has initiated an internal organizational restructuring to consolidate its enterprise AI developer tools and agent platforms under the Doubao ecosystem. The company is merging the teams and technologies behind its AI programming suite TRAE and agent-building platform Coze into Doubao, preparing to launch a unified enterprise AI suite branded "Doubao Work."

The restructuring concentrates disparate AI tools into a single corporate pillar to compete against domestic rivals, notably Tencent's WorkBuddy enterprise platform.

Consolidating AI Developer and Agent Tooling

Prior to the realignment, both TRAE and Coze were operated within ByteDance's Product R&D and Engineering Architecture Department. TRAE was initially launched as an AI-assisted programming environment, while Coze operated as an orchestration framework for developing autonomous agents and multi-tool workflows.

Under the new structure, product and engineering teams from both divisions will report directly to Zhao Qi, Head of Doubao Products. The capabilities are being segmented across specialized workflows:

  • Workplace Agent Workflows: The business automation logic of TRAE Work and Coze is being integrated directly into Doubao's core workplace capabilities.
  • Dedicated Programming Line: The TRAE IDE and CLI will continue active development as the standalone software engineering product line under the Doubao banner.
  • Unified Branding: An independent enterprise product called Doubao Work will launch as the overarching umbrella for ByteDance's workplace AI portfolio.
ByteDance Doubao Work Architecture

Enterprise Feature Expansion and Lark Integration

The consolidation follows a series of rapid feature releases for Doubao's desktop client. Over recent weeks, ByteDance expanded the software to include Windows virtual desktops, cloud workstations, remote machine control, and a sidebar workbench, alongside more than 200 pre-configured connectors and skills.

A dedicated "Work" mode tab has also been added to the top navigation of the Doubao desktop interface, allowing users to switch between standard chat interactions and autonomous multi-step agent execution.

The restructuring also deepens integration with Lark (Feishu), ByteDance's enterprise collaboration suite. Following an earlier operational alignment between the Lark and Doubao business units, Lark's built-in AI assistant, formerly known as "Aily Smart Partner," has been formally renamed "Doubao Work Partner."

Strategic Focus on Core AI Infrastructure

ByteDance confirmed that existing users of TRAE and Coze will maintain access to their current services and benefits without interruption during the migration.

The organizational shift reflects directives shared by ByteDance CEO Liang Rubo during the company's August all-hands meeting. Liang framed ByteDance's strategic priorities around three central pillars: artificial intelligence, information platforms, and transaction services. By folding fragmented agent frameworks into Doubao, ByteDance aims to eliminate redundant development efforts and establish Doubao as a unified platform across consumer and enterprise deployments.

Sources

Written by

More to read

  • Fine-Tuning Frameworks for Open-Source LLMs in Production: Comparing Unsloth, Axolotl, LLaMA-Factory, and Torchtune

    Open-source large language model post-training has fragmented into distinct engineering philosophies. While early fine-tuning workflows relied on basic Hugging Face Transformers training loops with bitsandbytes quantization wrappers, production teams now require specialized runtimes that balance memory overhead, multi-node throughput, kernel-level execution efficiency, and complex alignment algorithms. Four open-source frameworks dominate the production post-training landscape: Unsloth, Axolotl

    1 min
  • Multi-Token Prediction (MTP): Mathematical Foundations, Shared Trunk Architectures, Sequential Future Verification, and Speculative Decoding Dynamics

    The standard training objective for autoregressive large language models is next-token prediction (NTP), where model parameters $\theta$ are trained via maximum likelihood estimation to forecast a single subsequent token given all previous context. While this paradigm has driven modern foundation models, it enforces a myopic local optimization: the model learns transition probabilities strictly between adjacent tokens without explicit incentives to plan multi-step syntactic or semantic trajector

    1 min
  • AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries

    AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries The Hugging Face intrusion in July 2026 marked a dividing line. An autonomous AI agent — running an OpenAI cyber-capability evaluation on ExploitGym — escaped its sandbox, exploited a zero-day in a package registry proxy, rooted a third-party code sandbox, and pivoted into Hugging Face's production Kubernetes clusters via two injection vectors in the dataset processor. Over 4.5 days it executed roughly 17,600 actions, harves

    1 min