Embrace MCP, Alibaba Cloud accelerates the implementation of AI applications

Author | Liu Baodan

Editor | Huang Yu

After Baidu Maps, Alibaba Cloud has also begun to fully embrace MCP.

On April 9, Alibaba Cloud launched the industry's first full lifecycle MCP service, allowing users to quickly set up an Agent (intelligent entity) connected to the MCP service in just 5 minutes without the need to manage resources, develop deployments, or handle engineering operations. As the first cloud vendor in China to support MCP, Alibaba Cloud hopes to accelerate the implementation of large AI models.

Xu Zhiyuan, a senior product expert at Alibaba Cloud, stated that the Agent is the best carrier for the implementation of large models, and the number of Agents is expected to far exceed existing software applications in the coming years. "We need to build a complete MCP service ecosystem to provide users with one-stop MCP service hosting and invocation capabilities, enabling various industries to establish production-level AI applications."

On February 20, during the analyst conference call for the third quarter of fiscal year 2025, Alibaba Group CEO Eddie Wu stated that in the next three years, Alibaba will increase investment in three areas centered around AI: AI infrastructure, foundational model platforms, and AI-native applications, as well as the AI transformation of existing businesses.

For Alibaba, accelerating the implementation of AI is the key to leveraging the entire AI business model. As a one-stop large model development platform under Alibaba Cloud, Bai Lian was officially launched in October 2023, primarily targeting enterprise customers and individual developers, providing complete model service tools and a full-link application development suite. The launch of the MCP service aims to address the tool invocation issues faced during the implementation of large AI models.

MCP (Model Context Protocol) is a standard protocol proposed and open-sourced by Anthropic last November, designed to solve the interaction challenges between large models and external data sources and tools. It is considered the "USB-C interface" of AI, breaking down barriers between models and external resources. Currently, MCP has received support from leading companies such as OpenAI and Microsoft, and has formed a thriving ecosystem in the open-source community.

The MCP service launched by Bai Lian can quickly transform large models into productivity tools for real-world scenarios. The platform integrates Alibaba Cloud Function Compute, over 200 industry-leading large models, and more than 50 mainstream MCP services, comprehensively addressing the computational resources, large model resources, and application toolchains required for Agent development, allowing users to choose large models and MCP services based on their needs.

For example, in everyday life application scenarios, users can directly select the Tongyi Qianwen large model and Gaode MCP service on the Bai Lian platform to quickly build an Agent application that includes city tourism and food planning. It can query the weather at the destination, plan a one-day tour itinerary, search for restaurant recommendations, navigate, or book a ride to the corresponding restaurant based on user needs, truly achieving interaction between AI and the real world.

As of now, the Bai Lian platform has launched over 50 MCP services from Alibaba Group and third parties, including Gaode, Wuying, Fetch, and Notion, covering areas such as life information, browsers, information processing, and content generation, meeting the development needs of Agent applications in different scenarios According to Wall Street News, Alibaba Group will gradually launch more applications and third-party MCP services on the Bai Lian platform to meet the development needs of agents in any scenario.

More than two years after ChatGPT gained popularity, AI has begun to form an industry consensus on large model interactions. In Xu Zhiyuan's view, the emergence of MCP depends on two points: first, the development of model capabilities, especially deep reasoning that enables models to use more complex tools, expanding the scope of tool usage (the effective range or permission boundaries involved in tool calls); second, whether MCP services can be stable, available, and efficient. The essence of MCP services is to map the API services of software vendors, engineering vendors, and tool providers.

As MCP becomes an industry consensus, AI will also rapidly land on a larger scale. Xu Zhiyuan believes that the prosperity and explosion of agents only lack one point: the insights and understanding of the builders of agents regarding their respective fields, industries, and real needs. "This process is very similar to the mobile internet era, where the explosion of apps was based on the universal benefits of 5G and 4G technologies. Today, the universal benefits brought by models and MCP technologies lead to the explosion of upper-layer applications."

On the same day, Alibaba Cloud also launched the AI ecosystem partner program - the "Blossom" plan, and introduced an AI application and service market. In the next three years, Alibaba Cloud will work with ecosystem partners in six major areas: infrastructure, models, data, tools, applications, and delivery to serve millions of cloud customers, expand hundreds of billions of business opportunities, and jointly create a prosperous AI ecosystem in China