How ByteDance Integrated TRAE Work into Doubao in 30 Days: The Underlying Logic

Timeline and Scope of the Integration

On August 24, ByteDance merged the TRAE and Coze (formerly "扣子") teams into the Doubao ecosystem; the next day Doubao Work was officially launched as a unified AI office product. The integration combined product and operations under Doubao product lead Zhao Qi, promising better collaboration and no impact on existing user rights.

The division of labor was explicit: TRAE Work, Coze and Doubao’s office capabilities were consolidated, while TRAE IDE and CLI continued as a programming product line under the Doubao brand. A month earlier, on July 30, the Feishu product team had already been merged into Doubao, and on August 6 CEO Liang Rubo emphasized AI, information platforms and transaction services as core businesses, positioning Doubao as the “backbone” product.

What TRAE Work Is

TRAE Work evolved from an IDE plugin in 2025 to an independent product called TRAE SOLO in March 2026, and was renamed TRAE Work on June 9 2026, expanding from a developer‑only tool to an AI workspace for developers, professionals and students. The team explained the rename as a shift from a developer‑centric “SOLO” to a broader “Work” that removes the boundary.

The product splits into two modes:

Work mode : targets non‑technical roles (product, operations, marketing) and uses natural language to create content, analyze data, and draft proposals.

Code mode : retains full programming capabilities, supporting complex development, debugging and code‑base management. The underlying execution engine, inherited from the SOLO agent architecture, keeps the same core logic—context understanding, tool scheduling, iterative task execution, and traceable result delivery—while expanding the processing target from code files (.ts, .py) to office files (.pptx, .csv, documents).

Supporting features include a unified workspace that avoids repeated uploads, an extensible skill system integrated with the Feishu ecosystem via the MCP protocol, cloud‑run tasks that continue when the local client is closed, mobile progress monitoring, a Wiki generator that can index over 1,000 Markdown pages, and a Design mode that converts ideas into interactive prototypes.

Shortcomings are also noted: a free‑quota tightening in April 2026 reduced the user experience, remote control is limited to ByteDance’s own mobile app, and deep integration with document and IM ecosystems is still maturing.

Why TRAE Work Did Not Become the Flagship Against WorkBuddy

Market data from the iResearch Q2 2026 report shows 17 mainstream desktop AI office agents achieving over 60 million monthly visits, with Tencent WorkBuddy leading at 20.97 million visits, surpassing the combined second and third places. ByteDance’s TRAE IDE domestic version logged 12.79 million visits, while Alibaba’s QoderWork recorded 7.88 million.

Four reasons explain the gap:

The core user base of TRAE remains developers; the Work mode’s broader story has not translated into office‑side traffic.

TRAE runs only ByteDance’s own models, creating a relatively closed ecosystem, whereas WorkBuddy aggregates models such as DeepSeek, Mixed‑yuan, GLM and Kimi, allowing users to switch via points.

The early tightening of free quotas lowered the experience just as competitors were subsidizing usage.

Integration depth with documents and instant messaging is weaker, limiting the contextual power needed for office agents.

The fundamental issue, however, is the lack of enterprise‑level context. TRAE excels at execution but does not inherently understand a company’s internal data. WorkBuddy leverages the WeChat ecosystem for context, while Doubao relies on Feishu’s ten‑year B‑to‑C context foundation. Consequently, the optimal strategy is to extract TRAE’s execution capability and feed it into a product that already has a national‑scale entry point.

Three‑Step Strategic Play: Feishu Supplies Context, TRAE Supplies Execution, Coze Supplies Ecosystem

ByteDance’s three moves are not mere team mergers but targeted gap‑filling:

Feishu adds context : On July 30 the Feishu team merged into Doubao, bringing a decade of documents, meetings, multi‑dimensional sheets, approvals, knowledge bases, and a full identity‑and‑permission system.

TRAE adds execution : On August 24 the TRAE team contributed the SOLO agent architecture—task decomposition, tool scheduling, continuous iteration, traceable delivery—and the experience of extending processing from code files to office files.

Coze adds ecosystem : Also on August 24 Coze contributed an agent development platform and skill ecosystem, launching over 200 skills and connectors and a “work partner” system within a week.

Between these steps, a rapid rollout occurred: remote‑control for mobile on August 17, virtual desktop on August 18, dual‑mode cloud‑local switching on August 19, side‑panel workspace on August 20, followed by the skill store, connectors and the rebranded “Doubao Work Partner”.

Resulting in a consolidated B‑to‑B offering: Doubao’s near‑300 million monthly active users serve as the entry point, Feishu provides the contextual backbone, TRAE supplies execution and programming capabilities, and Coze delivers the intelligent‑agent ecosystem.

Retained Components: TRAE IDE and CLI as a Dedicated Programming Line

Importantly, TRAE was not eliminated; it was split. TRAE Work’s office functions moved into Doubao Work, while TRAE IDE and CLI remain under the Doubao brand as a dedicated programming product line. TRAE IDE continues to embed deeply in software development workflows, retaining IDE mode and legacy SOLO mode for professional developers, whereas TRAE Work pursues a broader, cross‑role approach.

This split reflects a realistic market segmentation: programming demands deep, specialized tooling, while office productivity requires a unified, low‑friction entry point. Merging both would dilute depth, akin to a tool that must both write code for engineers and design PPTs for marketers.

Industry‑Wide Consolidation Toward a Unified Entry Point

ByteDance’s move mirrors actions by other major players. Tencent, in July, folded QClaw into the WorkBuddy cloud product team, keeping the two products operationally independent—a lightweight resource consolidation. Alibaba merged QoderWork, Wukong and MuleRun into the Qianwen office suite under DingTalk leadership, creating a “technology‑three‑in‑one” approach. ByteDance’s reorganization is the most thorough, absorbing Feishu, TRAE and Coze entirely into Doubao.

IDC China analyst Sun Zhenya explains that vendors consolidate to eliminate duplicate early‑stage investments and reduce procurement confusion for enterprises. Three concurrent signals drive this trend: rapid decline of technical barriers (the “agent” question is no longer decisive), rising cost of scattered resources, and a shift in competition focus from individual tools to the “first window” users open each day.

Being that first window grants control over enterprise data and capability orchestration, but traffic dominance does not guarantee revenue. Free usage lowers entry barriers without building a moat; true value lies in converting trial users into daily paid users and binding them to the ecosystem.

Looking ahead, the next 12‑18 months will be judged by product revenue, monthly active users, token consumption, task completion rates, and enterprise renewal rates—metrics that indicate whether “sheep‑flock” users become paying customers.