If Intelligent Agents Were to Hold a Grand Parade: Six Major Formations of Domestic AI Agent Platforms

AI Agent (智能体)

Platform Overview

Market Growth

The parade symbolizes military strength, as well as the embodiment of national will and technological prowess. On the battlefield of AI, a “parade” in the field of intelligent agents is quietly unfolding—not in the form of neatly marching formations, but rather a hundred schools of thought contending in technology and a thousand ships competing in application.

In recent years, AI Agent technology has made rapid advancements, with technology companies both domestically and internationally entering the fray, launching various distinctive intelligent agent platforms, presenting an unprecedented scene of prosperity. According to incomplete statistics, there are currently over 120 intelligent agent platforms emerging in the domestic market, each like different formations in a parade, with varying technical backgrounds and application focuses, yet all pointing towards a smarter, more autonomous future of human-machine collaboration.

Today, let us review these AI Agent platforms from the perspective of a “technological parade” and see how they reshape industries and define the future.

01

Commander of the Parade—Core Value of Intelligent Agent Platforms

Intelligent Agent Technology: A Leap from “Response” to “Action”

If traditional AI is like a “scholar well-versed in many subjects,” then AI Agents are “versatile practitioners who can think and execute.” Their core capabilities can be summarized in four points: autonomy, reactivity, sociality, and learning. They not only understand commands but can autonomously call tools, execute tasks, interact with the environment, and continuously optimize in the process.

“The ability to call tools is the key differentiator of AI Agents from traditional conversational AI. This is not only an upgrade in technical architecture but also a fundamental shift in the role of AI—from a supportive ‘advisor’ to a capable, closed-loop executing ‘intelligent employee.'”

For example, in planning a trip, traditional generative AI only provides informational references, while a mature travel agent would comprehensively consider your budget, preferences, and real-time weather, automatically calling ticket booking APIs, map services, and hotel systems to directly generate orders and itineraries. This integrated “thinking-executing” capability is the true disruptive nature of intelligent agent technology.

(For an introduction to the concept of Agent intelligence, see: Dissecting the “Five Organs and Six Bowels” of Intelligent Agents: Perception, Decision-Making, Planning and Execution, Memory, Learning)

Platform Role: Low-Code + Multi-Modal + Industry Customization, the Triad of Empowerment

Intelligent agent platforms play a dual role in this technological revolution as both “training grounds and equipment depots.” They reduce development barriers and enhance interaction experiences through three core capabilities:

• Low-Code Development:Allows business personnel to quickly build applications through a drag-and-drop interface, greatly unleashing AI productivity;

• Multi-Modal Interaction:Integrates voice, gesture, text, and other multi-channel interactions, making human-machine communication as natural as interpersonal dialogue;

• Industry Customization:Dives deep into vertical fields such as healthcare, finance, and manufacturing, providing plug-and-play professional knowledge and business process support.

The competition among intelligent agent platforms is essentially a competition for “ecological adaptability.” The future winners may not necessarily be the ones with the most advanced technology, but those who can integrate into industry scenarios the fastest and most deeply.

Market Outlook: $285 Billion, A Technology Arms Race Just Beginning

According to Roots Analysis, it is predicted that by 2035, the global AI Agent market size will exceed $216.8 billion, with a compound annual growth rate of 40.15%. Behind this figure are two core driving forces:

The maturity point of technology has arrived: the generalization ability of large models, reduced inference costs, and improved reliability of tool calls;

Enterprise digitalization is entering deep waters: traditional automation bottlenecks are emerging, and AI Agents are becoming a new engine for business process reconstruction.

We must also be clear that market competition is heating up rapidly. Whether it is internet giants, cloud computing vendors, or startups, all are accelerating their layouts, striving to seize ecological positions. Technology can be open-sourced, but scenarios, data, and industry know-how are the real moats.

02

Review of Six Major Formations of Intelligent Agents

Internet Giants Formation: The Strong Ones in Ecological Integration

In the “parade” of intelligent agents, the formation of internet giants is undoubtedly the most eye-catching presence. Alibaba Cloud’s Bai Lian, Tencent Cloud’s Intelligent Agent, and Baidu’s Wenxin Intelligent Agent lead the way, occupying an important position in the field of intelligent agents due to their strong ecological integration capabilities.

Alibaba Cloud Bai Lianis a one-stop AI model training and application development platform launched by Alibaba Cloud, based on the Tongyi foundational large model, providing a complete toolchain from model construction to deployment for enterprise clients, ISV developers/technical personnel, and researchers. It supports integration of exclusive large model capabilities through API services and can connect its own business APIs into business workflows. Its core advantage lies in providing pre-trained models in multiple fields such as image recognition, speech processing, and natural language understanding, while also supporting custom model training to meet personalized needs, adapting to various industry scenarios such as finance, healthcare, e-commerce, and education through a user-friendly console interface and full-link model services (covering data management, training, deployment, etc.).

Official website: bailian.console.aliyun.com/#/home

Tencent Cloud Intelligent Agent Development PlatformAs Tencent Cloud’s AI development tool for enterprises and individual users, it relies on the multi-regional deployment capabilities of Tencent Cloud’s CVM cloud servers (covering multiple domestic and international nodes) and high-performance computing support (such as the second-generation CVM using Intel Broadwell CPU, DDR4 memory, with a peak concurrency of 300,000), providing a stable and reliable underlying operating environment. Its core highlight is the zero-code rapid creation of multi-agent collaborative interaction systems, supporting integration with mainstream databases and Tencent Docs as data sources, achieving flexible configuration of complex scenarios and more accurate and efficient dialogue responses; it also features enterprise-level security guarantees and high availability, with data reliability reaching four nines to eight nines, meeting the rigorous demands of B-end clients.

Official website: tencentcloud.com/zh/products/tcadp

Baidu Wenxin Intelligent Agent PlatformWith Baidu’s deep accumulation of artificial intelligence technology and a vast knowledge graph, it has garnered significant attention. The platform offers diverse development modes, allowing developers to create highly personalized intelligent agents using zero-code natural language creation features or one-click configuration of digital personas. In terms of knowledge and data processing, it relies on Baidu’s knowledge graph for knowledge enhancement, possesses leading domestic data set access capabilities, and aggregates a wealth of tool plugins, connecting multiple scenarios and device distribution channels.

Official website: agents.baidu.com/center

Cloud Computing Service Providers Formation: Ensuring Elastic Computing Power and Secure Deployment

The cloud computing service providers formation also holds an important position in the intelligent agent “parade,” with Huawei Cloud ModelArts, JD Cloud JoyAgent, and 360 Zhiyu leading the way, showcasing unique advantages in elastic computing power and secure deployment.

Huawei Cloud ModelArts Versatileis an enterprise-level agent platform launched by Huawei Cloud in 2025, aimed at helping enterprises and developers quickly build and deploy professional, efficient AI intelligent agent applications. The platform comes with rich business scenario experience templates, supporting low-code or even no-code intelligent agent development methods, significantly lowering the technical barriers for AI applications, achieving efficiency improvements of “minute-level development, day-level results.” ModelArts Versatile is derived from over 20 years of business process and practical experience within Huawei, possessing the high yield, professionalism, proactivity, and security characteristics required for enterprise-level applications, and can be widely applied in government affairs, healthcare, finance, manufacturing, travel, and other industries, accelerating the intelligent upgrade of enterprises.

Official website: huaweicloud.com/product/pangu/modelartsstudio

JD Cloud JoyAgentis the industry’s first 100% open-source enterprise-level intelligent agent solution, providing a complete set of product-level open-source components, including front-end, back-end, framework, engine, and core sub-agents. The platform is based on the practical experience of over 20,000 intelligent agents within JD, featuring a multi-agent collaborative engine that can efficiently handle complex tasks through discussion or voting mechanisms, achieving an accuracy rate of over 75% in the GAIA benchmark test, surpassing several well-known products in the industry. JoyAgent supports local independent deployment without relying on a specific cloud platform and comes pre-equipped with various templates and tools to help enterprises quickly build and deploy AI intelligent agent applications, bridging the “last mile” of AI implementation.

Official website: jdcloud.com/cn/products/agents

360 Zhiyuis an intelligent business collaboration platform under the 360 Group, focusing on the large government and enterprise market, formerly known as “360 Zhiyu.” It provides a secure and controllable AI collaborative office solution for large organizations such as government, military, and state-owned enterprises through its product architecture of “two bases” (AI service base and communication connection base) and “five capability” modules (intelligent agents, business integration, application generation, task collaboration, and security). The platform effectively integrates dispersed business systems, breaking down information silos, achieving intelligent collaboration across systems and organizations, and includes a super AI assistant “Zhixiaoyu” to provide voice interaction, intelligent summarization, document processing, and other capabilities, with applications in over 20 industries such as public security, finance, and manufacturing, helping government and enterprise clients improve work efficiency and safely introduce AI technology.

Official website: ccwork.com.cn

AI and Large Model Vendors Formation: Pioneers of Technological Innovation

The AI and large model vendors formation showcases unique charm through cutting-edge technological innovation in the intelligent agent “parade.” MiniMax Xingye Intelligent Agent, Zhipu Qingyan, and Mianbi Intelligent ChatDev represent the forefront of technological innovation.

MiniMax Xingye Intelligent Agentis an immersive virtual AI social application developed by the large model unicorn MiniMax, integrating text, sound, and visual multi-modal technologies, focusing on personalized intelligent agent creation and ultra-realistic interactive experiences. Users can freely customize the agent’s appearance, voice, persona, and skills, achieving smooth and natural interactions through real-time voice conversations, with its hyper-human-like voice and dynamic expressions (such as smiling, shyness, etc.) further bridging the emotional distance between AI and humans. As the user’s “AI partner,” Xingye covers diverse scenarios such as entertainment, work, and learning, supporting multi-role switching and memory functions, meeting both knowledge Q&A and interactive chat needs, while providing thoughtful companionship through emotional recognition, becoming a new choice for millions of users in daily communication.

Official website: xingyeai.com

Zhipu Qingyanis a generative AI assistant launched by Zhipu AI based on its self-developed ChatGLM-4 large model, supporting bilingual interaction in Chinese and English, with core capabilities in general Q&A, creative writing, and code generation, and open multi-modal generation functions such as text-to-image and text-to-video. Its feature lies in the “intelligent agent” customization function, allowing users to create intelligent agents in roles such as teachers and artists, simplifying technical barriers through prompt engineering, and combining knowledge base uploads for specialized task processing, such as thesis summarization and code debugging. This product supports multi-end use, covering iOS, Android, and PC, providing practical tools such as video calls, online searches, and one-click PPT generation, widely applied in office automation, creative design, and learning assistance scenarios, becoming a smart life and work partner for users due to its ease of use and diverse functionalities.

Official website: chatglm.cn/main/toolsCenter

Mianbi Intelligent ChatDevis a full-process automated software development framework developed by Mianbi Intelligent in collaboration with Tsinghua University’s NLP laboratory, centered on collective intelligence technology, simulating a virtual software company’s multi-role collaboration (such as CEO, CTO, programmer, etc.) to achieve full-process automation from requirement analysis to code generation. This framework allows users to input requirements in natural language, completing software development in as fast as 3 minutes (169 seconds) at an average cost of less than 3 RMB, and integrates Git version control and Human-Agent interaction features, allowing users to supervise and adjust the development process in real-time. As an open-source project with over 12,000 stars on GitHub, ChatDev has launched a SaaS version, covering diverse software development scenarios such as Snake game and efficiency toolset, using a “1 person + multiple AI agents” model to lower the entrepreneurial threshold, becoming an efficient tool connecting creativity and productivity.

Official website: modelbest.cn

Traditional Software Transformation Formation: A Model of Old Veterans with New Intelligence

The traditional software transformation formation showcases the charm of old veterans with new intelligence in the intelligent agent “parade.” Kingdee Cangqiong Agent, Yonyou YonGPT, and Lenovo Baiying lead the way, providing strong support for enterprise digital transformation by deeply integrating ERP with AI.

Kingdee Cangqiong Agent Platform 2.0is the core technical foundation for Kingdee’s AI transformation for enterprises, aimed at lowering the technical barriers for enterprises to apply AI, accelerating the widespread development of intelligent agents through rich templates and tools, deep SaaS connections, enterprise-level security guarantees, and open technical standards. Its core functions include supporting multi-scenario AI application construction, having created over 100 AI scenarios based on the Cangqiong platform and co-created over 1,000 applications with clients, covering core management scenarios such as financial analysis (e.g., King Key Financial Report Intelligent Agent), data queries (ChatBI Intelligent Agent), travel management, recruitment, etc. The platform emphasizes “practical enterprise management AI” through a three-in-one architecture of “entry-application-platform” (e.g., Xiaoyun Assistant as the AI entry scheduling multi-agent collaboration), achieving task decomposition, process automation, and cross-scenario collaboration, helping nearly a hundred enterprises such as Hisense and Wens Food Group improve the efficiency of their digital transformation and promote changes in enterprise organizational models and work methods.

Official website: kingdee.com/topic/kccs2025/

Yonyou YonGPTis an enterprise service large model that integrates professional knowledge from multiple fields and industry business know-how, based on general large model pre-training and fine-tuning, covering ten major field scenarios such as finance, human resources, supply chain, and procurement, achieving intelligent data analysis and automated process generation by combining Yonyou’s product capabilities in various fields, providing intelligent decision support and efficient business processing for enterprises. Its core advantages lie in the rich accumulation of enterprise service experience, comprehensive solution coverage, and strong support from a powerful R&D team, having empowered dozens of application scenarios such as intelligent inventory optimization, intelligent recruitment, and code generation, and addressing core needs such as enterprise interaction flexibility, data security and privacy, and task controllability through technologies such as precise identification of intent slots, enterprise private domain knowledge Q&A, and decomposition of complex task execution chains.

Official website: yonyou.com/subject/gd-YonGPT

Lenovo Baiyingis Lenovo’s intelligent service system for small and medium-sized enterprises, with the core carrier being the Lenovo Baiying intelligent agent (upgraded to version 2.0), integrating three core applications: AI marketing, AI office, and AI services, through five technical capabilities such as multi-modal interaction and multi-agent collaboration, helping enterprises achieve cost reduction, efficiency enhancement, and open-source. AI marketing relies on the Lemi engine to improve customer acquisition efficiency several times, AI office integrates cloud collaboration functions to double efficiency, and AI services reduce IT costs by about 50% through intelligent operations; the 2.0 version further upgrades to L3-level intelligent agents, adding features such as IT code resolution (end-to-end solution for over 200 IT issues), super employees (task implementation driven by natural language), and zero-threshold marketing design (efficiency improvement by ten times), supporting multi-end linkage. The platform offers a six-month free public test, covering the panoramic business scenarios of small and medium-sized enterprises, becoming a key tool for promoting digital transformation.

Official website: baiying.lenovo.com

Process Management and Automation Formation: Aiding Efficiency Enhancement

The process management and automation formation in the intelligent agent “parade” is dedicated to achieving intelligent management and automated execution of enterprise processes through the integration of AI large models as new digital employees. Shizai Agent Intelligent Agent, Dezhuzhizhi Intelligent Agent Platform, and Yujv AI Intelligent Agent Platform represent strong capabilities.

Shizai Agent Intelligent Agentis the world’s first general intelligent agent developed by Shizai Intelligent, integrating third-generation RPA technology, large language models, and multi-modal interaction capabilities, possessing three core abilities: human-level reasoning, infinite linking, and scene adaptation, with a task decomposition accuracy rate surpassing GPT-4, achieving technological landing two years ahead of competitors. As a “digital life form,” it empowers government and enterprise users through four major functions: supporting natural language interaction for precise understanding of needs, providing personalized content generation (such as generating long reports), conversational data analysis (capturing and processing multi-source data to gain insights), and full-process automated execution (such as project management decomposing tasks and connecting business systems), addressing the defect of large models that “can think but cannot do.” Its advantages lie in high intelligence (deeply processing complex scenarios at the core of business), strong language understanding and generation capabilities (lowering usage barriers), autonomous task decomposition and execution (such as full-process planning for software development projects), and real-time learning and adaptation capabilities, having helped over 4,000 enterprises achieve quality improvement, cost reduction, and efficiency enhancement, opening a new mode of human-machine collaboration.

Official website: ai-indeed.com/products/agentRpa

Dezhuzhizhi Intelligent Agent Platform(under Zhongguancun Science and Technology Financial Group) is an AI intelligent interaction and knowledge management platform for enterprises, with core advantages in strong knowledge base construction capabilities and multi-modal interaction support. Its architecture supports parsing various formats, with document knowledge base and FAQ knowledge base management functions, and provides an out-of-the-box RAG (Retrieval-Augmented Generation) framework, efficiently processing text, voice, and other information. Technically, it integrates speech recognition and synthesis capabilities, achieving full-process intelligent processing from voice consultation to text analysis, suitable for customer service dialogues, meeting records, and other scenarios. Application scenarios cover enterprise intelligent customer service, knowledge base Q&A systems, and automated customer consultation, enhancing service quality and operational efficiency for enterprises by optimizing information retrieval and response efficiency.

Official website: 51ima.com/agentflow

Yujv AI Intelligent Agent Platformis an intelligent tool integrating advanced natural language processing technology, with core functions covering text generation and analysis (such as intelligent writing, sentiment analysis), language translation and conversion (multi-language mutual translation, voice/text conversion), dialogue interaction (intelligent Q&A, chatbots), and security and privacy protection, supporting text proofreading, keyword extraction, and cross-platform synchronization. Its feature lies in open ecological integration, allowing access to third-party intelligent agents (such as Coze, Tencent Yuanqi), and achieving complex task automation through AI process templates and over 20 action configurations, such as generating code comments and technical documents using AutoGen Studio, or quickly achieving application internationalization through multi-language translation functions. The platform is suitable for enterprise customer service, content creation, development assistance, etc., ensuring user information security through data encryption and privacy settings, while enhancing user experience through personalized recommendations and cross-device synchronization, becoming an efficient assistant in the field of language processing and intelligent interaction.

Official website: yuju-ai.com

Open Source Pioneer Formation: Leaders in Technology Sharing

The open-source pioneer formation promotes technological innovation and sharing through open code in the intelligent agent “parade,” providing new momentum for the development of intelligent agents. Dify, Yangmei Intelligent Application Development and Deployment Platform, and Kaiwu Intelligent Agent Middle Platform lead the way, showcasing the charm of open source.

Difyis an open-source large language model application development platform that supports hundreds of models such as GPT, Mistral, and Llama3, with a declarative development environment (defining applications through YAML), modular design, and LLMOps functionality (monitoring and optimizing application performance) as its core, simplifying the complex AI application development process, particularly suitable for deep customization or enterprise-level deployment scenarios. Its advantages lie in flexibility and scalability, supporting self-hosting and seamless integration with existing enterprise systems, meeting data security and compliance needs, while relying on an active open-source community to provide rich templates and collaboration opportunities (such as Workflow visualization). However, the platform has a higher learning curve, requiring technical background for model integration and configuration, and the domestic ecosystem and plugin support are weaker than Coze. Applicable scenarios include building enterprise-level LLM infrastructure, private deployment, and developer-led complex AI application development.

Official website: dify-china.com

Yangmei Industrial Platformis an industrial intelligent agent development and deployment platform independently developed by the Beijing Industrial Software Industry Innovation Center, adhering to the mission of “empowering with intelligence, reshaping the industrial future,” with core value in encapsulating vertical field knowledge into reusable industrial intelligent agents (“Natto”), promoting deep application of AI in industrial scenarios. The platform has three core capabilities: first, providing a dual-mode intelligent agent development engine, supporting zero-code rapid creation of industrial apps, agents, and models for scenarios such as process optimization and quality inspection, significantly lowering the barriers for AI applications; second, supporting agile deployment and application, adapting to mainstream industrial resources such as PLC, DCS, CAD through private solutions, bridging the “last mile” of implementation; third, building an open industrial ecosystem, connecting enterprises and universities to achieve end-to-end matching of demand and solutions.

Official website: yangmeigongye.com

Kaiwu Intelligent Agentis a core functional module integrated into the Kaiwu Cloud “Kaiwu Zhihua” intelligent agent all-in-one machine, designed to help enterprises build large model native applications quickly and at low cost. Its core value lies in providing a multi-modal collaborative development environment, seamlessly integrating various industry intelligent application building scenarios, supporting collaboration with tools such as Kaiwu AI Assistant and ChatBI data intelligent analysis platform, achieving full-process intelligence from intelligent Q&A, professional content generation to data management analysis. As a standard capability of the Kaiwu Zhihua product matrix, Kaiwu Intelligent Agent ensures high-precision reasoning and low-latency response through the native integration of full-blooded FP8 precision lossless version of DeepSeek-V3/R1 large model, while supporting cross-scenario business collaboration and decision-making efficiency improvement, providing secure and controllable intelligent application development and operational support for enterprise private deployment.

Official website: kaipuyun.cn

This “parade” showcases not only the accumulation of technology but also the roadmap of the collision between AI and application scenarios in deep waters. Currently, the development of intelligent agents is still in the “early explosion phase”; although there are many platforms, signs of homogenized competition are already emerging. The next stage of competition will shift from “model capability” to the triathlon of “engineering, scenario, and commercialization”.

Whoever can truly understand industry pain points, provide stable and reliable closed-loop implementations, and build an active developer ecosystem will be the one to ultimately succeed in this $200 billion intelligent revolution.

If Intelligent Agents Were to Hold a Grand Parade: Six Major Formations of Domestic AI Agent Platforms

© 2025. Intelligence Indeed