---
id: "lib-2026-11-04-coze-enterprise-and-tencent-yuanqi-shut-down"
type: "library"
title: "Coze Enterprise and Tencent Yuanqi shut down"
title_original: "字节扣子、腾讯元器接连下线：大厂AI产品洗牌清退潮"
lang: "ZH"
org: "Tech星球 (on 36Kr)"
source_type: "Practitioner post"
date: "2026-11-04"
week: "2026-W45"
url: "https://nosov.ai/library/lib-2026-11-04-coze-enterprise-and-tencent-yuanqi-shut-down/"
original_url: "https://www.36kr.com/p/4055286987838595"
tier: "B"
tier_name: "Useful, limited"
tier_reason: "Tech Xingqiu (Ren Xueyun, republished by 36Kr) compiles dated shutdown and merger announcements across Chinese big tech and AI start-ups with IDC and regulator figures; useful as a dated register, but the count is the outlet's own incomplete tally and the analysis is commentary."
summary_basis: "full-text"
original_minutes: "10"
tags: ["ai technology", "business transformation", "agent platforms", "vendor risk"]
---

# Coze Enterprise and Tencent Yuanqi shut down

Original title: 字节扣子、腾讯元器接连下线：大厂AI产品洗牌清退潮

Tier B (Useful, limited): Tech Xingqiu (Ren Xueyun, republished by 36Kr) compiles dated shutdown and merger announcements across Chinese big tech and AI start-ups with IDC and regulator figures; useful as a dated register, but the count is the outlet's own incomplete tally and the analysis is commentary.

Summary based on the full text.

## In short

A 4 November 2026 article from Tech Xingqiu (Tech星球), republished by 36Kr. ByteDance's Coze Enterprise stops new sales on 30 November 2026 and closes by 31 December 2027; Tencent's Yuanqi agent platform closes on 9 November 2026. The outlet counts 38 AI products shut or merged at Alibaba, ByteDance, Tencent, Baidu and start-ups including Moonshot and MiniMax. It names three causes: general models absorbing single-feature tools, compliance rules on human-like companions, and business models where usage drives losses.

## Key numbers

- 38: Chinese AI products shut down or merged, by Tech Xingqiu's incomplete count; 25 of them at big tech firms in H2 2026
- 19.3% of RMB 250 million: Coze's share of China's 2025 public-cloud agent-development-platform market, and the market's total size (IDC)
- over 8 million: user-built characters affected when Doubao, Qwen and Yuanbao removed custom-agent features in mid-July
- 302 to 1,112: registered generative AI services in China, end of 2024 vs 31 August 2026

## What it reports

- Coze Enterprise launched in H2 2025 and is being withdrawn after a little over a year; Yuanqi launched in May 2024. A self-described former Coze employee says workflow tools were displaced by models that can plan on their own.
- Tencent QClaw (remote phone control of a PC) launched March 2026 with over 100,000 sign-ups and was announced for shutdown in September, nine months after launch.
- Big-tech consolidation in H2 2026, 25 products: QClaw into WorkBuddy (20 July); Feishu and Doubao product teams merged (30 July); three Alibaba agent products into Qwen Office (3 August); TRAE and Coze teams into Doubao (24 August).
- Start-ups cut side projects: Moonshot dropped Ohai and Noisee for Kimi; StepFun closed Maopaoya (6 million+ downloads) on 5 November 2025; 01.AI closed its public model platform on 3 August 2026 and external API on 3 September to focus on enterprise work.
- IDC: Coze led China's 2025 public-cloud agent-development-platform market with 19.3%, in a market of only RMB 250 million.
- Rules on human-like interactive AI services took effect 15 July 2026, with "emotional dependence" a red line. In mid-July Doubao, Qwen and Yuanbao removed self-built agents, affecting over 8 million characters.
- Generative AI services registered in China grew from 302 at the end of 2024 to 1,112 by 31 August 2026, with another 731 apps or features registered.

## How they got it

Compilation of public shutdown notices and company announcements, plus IDC market data and regulator registration counts. The product tally is the outlet's own and described as incomplete.

## Limits

- The 38-product count is incomplete and mixes full shutdowns with mergers and rebrands.
- Causes are the author's analysis; no usage or revenue data for the closed products.
- One quote is from an unnamed former employee.

## Why it matters for companies

Agent-building platforms from large vendors can be withdrawn within a year, so anything built on them needs an export path and a migration plan. The IDC figure shows how small the standalone market for these platforms was, even for the leader. The pattern, single-purpose tools folded into a vendor's main assistant, argues for building around portable assets (prompts, data, tests) rather than a specific platform's workflow format.

## Key passages in translation

> Coze这种工作流产品不是被AI淘汰的，而是被"模型自己会规划"淘汰的。

Workflow products like Coze were not eliminated by AI; they were eliminated by models that can plan for themselves.

> 份额第一背后的现实是，赛道整体市场空间局限。

The reality behind the top market share is that the segment as a whole is small.

> 一款独立AI产品想要站稳脚跟，必须落到高频场景、拥有稳定付费群体，能够交付真实业务结果，而不只是展示模型能力。

A standalone AI product that wants to hold its ground must land in a high-frequency use case, have a stable paying user base and deliver real business results, not just show off model capability.

Original: https://www.36kr.com/p/4055286987838595