ByteDance Folds Trae and Coze Into Doubao, Preps 'Doubao Work' to Battle Tencent's WorkBuddy

ByteDance is executing one of the most consequential AI product consolidations in China this year. According to sources cited by Bloomberg on August 24, 2026, the company is merging its Trae coding platform and Coze agent-building tool into Doubao, its flagship AI super-app — and plans to launch a new office-productivity agent called Doubao Work, aimed squarely at Tencent’s surprise breakout hit, WorkBuddy.

The move ends the era of ByteDance running parallel AI product bets. Trae (the AI-native IDE that challenged Cursor and GitHub Copilot in the Chinese market) and Coze (the no-code agent platform that let anyone build and monetize chatbots, also operated internationally) will stop existing as standalone destinations and become capabilities inside Doubao. The message is unambiguous: ByteDance’s AI strategy now has a single front door, and everything feeds the funnel.

Why Consolidation, Why Now

Three pressures converged to force this decision.

First, Doubao’s scale demands it. Launched in August 2023, Doubao grew into China’s most widely used AI assistant, passing 155 million users by early 2026 and reportedly clearing 200 million as the Seed 2.0 model family rolled out. When one app owns that distribution, maintaining separate brands for coding and agent-building becomes an expensive redundancy. Folding Trae and Coze in gives Doubao the two capabilities users were leaving the app to find elsewhere — and gives ByteDance a single place to measure engagement, upsell subscriptions, and harvest training signals.

Second, Tencent forced the issue. WorkBuddy, Tencent’s desktop AI office agent, became China’s most popular AI workplace assistant within months of launch — a rare product win for Tencent after its stock rout earlier this year, and one significant enough that Bloomberg reported it “sparks AI turnaround hopes” ahead of Tencent’s earnings. From a single natural-language instruction, WorkBuddy plans, decomposes, and executes multi-step work across documents, spreadsheets, presentations, mail, meetings, and knowledge bases. It reads and writes Office files, coordinates multiple sub-agents, and ships “ready-to-use outputs” rather than suggestions. ByteDance could not let Tencent own the office-agent narrative unopposed.

Third, the super-app logic is winning in China. Where Western labs ship standalone apps and API products, the Chinese market has repeatedly shown that AI capabilities get their maximum leverage embedded in a super-app with payment, social, and content rails already attached. Doubao 2.0’s February relaunch as a “full agentic system” — plus the December debut of the Doubao Phone Assistant on ZTE hardware, which sold out 30,000 units in a day — established the playbook. Doubao Work is the office-tier extension of the same thesis.

What Doubao Work Will Face

Tencent’s WorkBuddy is a formidable benchmark. It integrates with IM, docs, mail, meetings, and knowledge bases in one desktop workstation; handles Word, Excel, PPT, PDF and image files natively; runs specialized “AI Expert” agents that can be coordinated in parallel; and has even shipped on HarmonyOS PCs as that platform’s first desktop office AI agent. Crucially, it benefits from being woven into Tencent’s enterprise ecosystem — WeChat Work (WeCom) distribution, Tencent Docs, and meeting infrastructure.

ByteDance’s counter is its consumer-side gravity. Doubao’s hundreds of millions of users, the recommendation-engine DNA inherited from Douyin, and the Flow division’s product velocity give Doubao Work instant distribution that no startup could buy. The Trae merger means Doubao Work arrives with credible developer-grade coding ability built in, and Coze brings a mature agent-authoring layer — meaning users won’t just use pre-built agents, they’ll be able to build their own inside the same app.

The Bigger Picture: AI’s Super-App Wars

This is not just a ByteDance-Tencent skirmish. It is the clearest evidence yet that 2026’s defining AI product battle in China is the agentic super-app: one interface where a general assistant, a coding agent, an agent-builder, and an office automator live together, share context, and bill through one subscription.

Every major Chinese player is converging on this shape from a different direction — Tencent from enterprise messaging, Alibaba from cloud and open models (Wan3.0’s document-to-video launched commercially just today, hours after its $10 billion share sale closed), and ByteDance from consumer attention. The consolidation also signals rough weather for standalone tools: if Trae and Coze — genuinely successful products — couldn’t survive as independent brands inside ByteDance, the standalone AI tool market in China is being judged by a brutally short leash. Global developers building vertical AI tools should take note: distribution beats differentiation when the platform players bundle.

What to Watch

1. Doubao Work’s launch window and feature parity with WorkBuddy — particularly whether it matches WorkBuddy’s file-format depth and multi-agent coordination on day one.
2. Coze’s international fate. Coze operated outside China too; the merger raises the question of whether the global agent-builder service gets folded into an international Doubao or quietly sunset.
3. Tencent’s response at its next earnings call — WorkBuddy momentum was already the bull case; a credible ByteDance rival changes the story.
4. Regulatory temperature. Beijing tightened rules on AI companion features in July; office agents touch enterprise data, and the Cyberspace Administration has signaled it is watching the agentic layer.

The consolidation reports are sourced, not officially announced — expect ByteDance to confirm at or near the Doubao Work launch. But the direction is clear: China’s AI race is consolidating into a handful of super-apps, and ByteDance just declared which hill it intends to die on.