On June 5, 2026, the Tencent Cloud AI Industry Application Conference was held in Beijing. Faced with the growing demand for AI efficiency among individuals and enterprises, Tencent Cloud launched the "Efficiency Agent Toolkit" — offering differentiated and personalized agent solutions tailored to different users and over 20 vertical scenarios.

For individual users, Tencent Cloud upgraded AI tools such as QClaw, WorkBuddy, Baoyuan, ima, and Tencent Docs, making efficiency easily accessible; for enterprise customers, the newly released WorkBuddy Enterprise AI workspace was introduced, along with upgrades to core products like ClawPro, the Tencent Cloud Intelligent Agent Development Platform (ADP), and Qidian Marketing Cloud, building a comprehensive service system covering the entire lifecycle.

image.png

Tang Daosheng and Yao Shunyu in dialogue

During the conference, Tang Daosheng, Senior Executive Vice President of Tencent Group and CEO of the Cloud and Smart Industry Business Group, and Yao Shunyu, Chief Scientist of Tencent AI, had a dialogue on stage, explaining Tencent's latest strategies and thinking in the second half of the AI era, and responding for the first time to the deep co-design approach between large models and product development.

Tang Daosheng stated that Tencent has always been committed to practical, user-friendly, and scalable AI, with the core experience being to be rooted in real scenarios. Real scenarios contain both user needs and data required for model iteration. Tencent’s rich product scenarios, interaction data, and ecosystem connections can provide high-quality context for models, as well as enable models to call tools and connect systems, truly completing the task cycle.

Yao Shunyu believes that the second half of the AI era has shifted from seeking methods to finding problems, with the core being the balanced triangle of Foundation (base model), Product (product), and Frontier (frontier model). The essential difference of LLMs lies in generalization, and data from different scenarios needs mutual empowerment. The rise of Agents makes environmental, evaluation, and reasoning coordination more critical than single model capabilities. Based on this, the new Hunyuan model continues to deeply co-design with products, improving model size while continuously enhancing performance through multi-product data feedback and systematic training.

Efficiency Agent Toolkit, Establishing a New Standard for AI Efficiency

In the second half of the AI era, agents have become an important engine for reshaping productivity. The Tencent Efficiency Agent Toolkit, centered around user needs, encapsulates agent capabilities into a multi-layered product system — from lightweight tools ready to use out of the box to enterprise-level platforms and industry solutions, meeting the usage needs of different people.

image.png

Tencent Launches Efficiency Agent Toolkit

For individual users, Tencent has launched several “ready-to-use” products. Among them, the local AI assistant QClaw introduces the “WeChat direct connection” mode and integrates with tools like Tencent Docs, Tencent Meetings, ima, and QQ Mail; the personal knowledge intelligent agent ima supports creating personalized agents and continuously understands users through a memory system; the personal AI assistant Baoyuan supports “Baoyuan Pi” to quickly access “Lobster”; the QQ browser also launched an accompanying agent service and jointly launched the industry's first college entrance exam consultant agent “Baoyuan College Entrance Exam Assistant” with Baoyuan.

For white-collar workers, developers, OPCs, and designers, Tencent has launched the “Buddy” family series of products to meet the needs of frequent productivity scenarios such as code development, document processing, and creative design. Among them, the AI intelligent desktop workspace WorkBuddy Personal Edition has seen 43 versions updated in three months and is currently the most popular efficiency agent tool in China based on DAU. CodeBuddy, the coding intelligent agent, first supports three forms: plugins, IDE, CLI, covering the entire development process including coding, review, testing, and operations. Tencent Miora and Ardot are designed for creative generation and interactive design scenarios to improve professional work efficiency.

For enterprise users, Tencent officially launched WorkBuddy Enterprise (WorkBuddy Enterprise) and the Office Agent Suite. WorkBuddy Enterprise provides three core capabilities: 7x24 expert digital employees, team management support through the “project” function, and an enterprise-level management console.

The Office Agent Suite integrates the capabilities of Tencent Docs, Tencent Cloud Disk, and Tencent LeXiang into the WorkBuddy workspace via One ID, connecting content creation, knowledge accumulation, and capability reuse throughout the entire chain. After integrating with WorkBuddy, users and agents can simultaneously write and collaborate on the same document; Tencent Meetings uses AI to accumulate meeting context and allows various agents to understand and invoke meeting content through interfaces like MCP, Skills, and CLI.

In addition, the enterprise agent management and operation platform ClawPro and the Tencent Cloud Agent Development Platform ADP were also upgraded. ADP 4.0 focuses on building an “enterprise-level AgentOps platform,” covering the entire lifecycle of agent construction, connection, distribution, and governance.

In vertical business scenarios, Tencent has packaged the underlying platform capabilities into deliverable solutions, creating products such as Qidian Marketing Cloud MAGIC Agent, Operation Intelligent Agent CloudQ, DatabaseClaw, DataBuddy, Tianyu Risk Control Agent, Education Intelligent Agent LearnBuddy, Tencent Health Open Platform, and Tencent Electronic Signature AI Contract Assistant.

Based on this new efficiency agent toolkit, Tencent has already implemented it internally. For example, CodeBuddy has reshaped the internal R&D pattern at Tencent, achieving AI across the entire coding, review, testing, and operations workflow, covering over 95% of engineers and reducing overall coding time by 40%. WorkBuddy enables human-AI collaboration in development, allowing a small team of a few people to complete the initial version in just two days, and the product maintains a release speed of two versions per day after launch. The fully upgraded Agent Infra has compressed the agent deployment and iteration cycle from quarters to weeks. Tencent hopes to use this set of tools to bridge the last gap between “product power” and “productivity.”

Upgrading the Cloud Infrastructure for Agents to Support Large-Scale Deployment

To support the large-scale application of Agents in real business scenarios, Tencent Cloud has comprehensively restructured its cloud product system, building full-stack capabilities covering models, inference services, Agent infrastructure, and ecological connections, continuously lowering the barriers for Agent development and application, and accelerating the entry of Agents into production practices.

In the model aspect, Tencent's Huyuan continues to optimize for Agent scenarios. The newly launched Hy3Preview focuses on task execution capability and cost-effectiveness and has been widely applied in products such as CodeBuddy, WorkBuddy, and Baoyuan. After integration with WorkBuddy, the first response speed increased by 54%, the average task completion time decreased by 47%, and the task success rate remained at 99.99%. On the global large model API calling platform OpenRouter, Hy3Preview has consistently ranked among the top in terms of token consumption and market share for multiple weeks.

Regarding inference services, leveraging diverse chip compatibility, tidal scheduling, and FlexKV distributed caching optimization technology, the Tencent Cloud large model service platform TokenHub has improved overall computing power utilization by 40% and cache hit rate by 85%, effectively reducing Agent inference costs. TokenHub platform integrates Huyuan and third-party mainstream models, supports new model Day0 access, unified protocol and Key management, and dynamically routes according to cost, performance, and smart effect, allowing different models to automatically collaborate within the same Agent based on scenarios. It is now available globally.

In terms of Agent runtime and security, Tencent Cloud has comprehensively upgraded Agent Runtime, building a native infrastructure covering operation, memory, storage, and secure access. Through elastic scheduling and on-demand scaling, 70% of idle computing power can be released; the Memory service helps Agents better understand users, reducing token consumption by 60% and increasing task success rate by 30% in long-task scenarios; Tencent Cloud also introduced zero-trust, zero-credential access mechanisms to ensure secure calls to enterprise systems and data by Agents, providing a foundational support for large-scale Agent deployment in production scenarios. In addition, Tencent Cloud launched CodeBuddy Security, allowing Agents to deeply participate in the discovery and governance of security vulnerabilities.

At the same time, to better support Agents accessing enterprise data and enhancing multimodal interaction capabilities, Tencent Cloud's database, big data, WAND audio-video media capabilities, EdgeOne Makers, and other products continue to be upgraded. In particular, in SaaS applications and connectors, Tencent has turned touchpoints used by billions of users, such as WeChat, Enterprise WeChat, Baoyuan, mini programs, and QQ browsers, into agent entrances and running interfaces. It has also skillized several Tencent products like Tencent Docs, QQ Browser, Tencent LeXiang, Tencent Map, Tencent Health, Wealth Management, and WeChat Reading, and listed them on SkillHub, allowing users to call product capabilities anytime through natural language. Currently, more than 70,000 skills have been accumulated on Tencent Cloud SkillHub.

Comprehensively Opening Up Tencent's Advantages, Building an Agent Industry Ecosystem Together

At the conference, Tencent also announced the comprehensive opening of agent capabilities. The second phase of the Tencent AI Co-Creation Program focuses on building the “Buddy AI Ecosystem Co-Creation Plan” around WorkBuddy. Among the 23 companies that appeared as the first round of co-creation partners, including Alistar, Anker Innovation, Baiwang Cloud, DeepSeek, Fanwei Network, KIMI, Lianchen Information, Li Weike Technology, China Unicom Cloud, MiniMax, Meituan, Ping An Bank, Tudaodao, SalesEase, Xiaoe Tong, Ctrip, Xinxing Optics, Future Intelligence, China Mobile, Zhongkang Technology, Zhipu, and ZTE Communications, they will leverage the Tencent Efficiency Agent Toolkit to upgrade and build applications, services, and products for the Agent era, providing convenient agent products for users in both software and hardware terminals.

Currently, the Tencent Efficiency Agent Toolkit has rich implementation practices in more than 20 industries, including healthcare, consumer electronics, finance, gaming, retail, and education, helping leading enterprises in these industries transform and build super teams of humans and AI collaboration.