One-click deployment of popular MCP Server

AI developers around the world are hotly discussing "MCP" (Model Context Protocol). Although this protocol did not attract widespread attention when it was released by Anthropic in 2024, Cursor announced the integration of MCP in early 2025, which quickly brought it into the developers' field of vision. The outbreak of Manus in March accelerated the popularity of MCP. On March 27, OpenAI officially announced that its Agent SDK fully supports the MCP protocol. This move marks that MCP will become the implementation standard in this field and will surely reshape the development and interaction methods of AI applications.

At present, most of the community's MCP Servers are deployed in local STDIO mode. Although this mode can support data interaction between basic model services and tools, and is acceptable for simple testing, it brings varying degrees of development complexity when it comes to specific development and debugging due to IO redirection. At the same time, with the increasing richness of AI scenarios, on the one hand, data access is no longer limited to local, and on the other hand, the business has requirements for architectural reliability. The MCP Server based on local deployment is bound to be unable to meet complex production needs. Therefore, MCP Server hosted on the cloud will become the mainstream trend in the future. Function Compute (FC) currently supports one-click hosting of open source MCP Server, and everyone is welcome to experience it.

• Attract more developers to participate in the MCP ecosystem construction

After the MCP protocol becomes the de facto standard, developers no longer need to write complex JSON Schema parameter descriptions for each Function, which greatly reduces the workload of repeated development. Through open source or third-party MCP Servers, developers can quickly share and reuse resources. For example, the Blender-MCP project allows users to convert natural language instructions into 3D modeling operations through the MCP protocol. The project received 5.4k stars within one week of being open source.

• SaaS service providers embrace MCP Server

With the popularity of MCP, SaaS service providers can reach new market and industry opportunities by integrating MCP Server. The Stdio and SSE standards of the MCP protocol require service and data providers to provide API access, and cloud hosting will be the best choice.

• Enterprise-level MCP Server requires security compliance and elastic scalability

MCP Server connects services/data to large models. If the data permission scope and sensitive data filtering of large models are not restricted, it will cause security and compliance risks to enterprises. Cloud hosting provides built-in security tools such as permission control, operation auditing, and user privacy protection, which greatly reduces the exposure of security risks and reduces compliance costs. At the same time, the popularity of MCP Server is a huge opportunity for service providers. Service providers will face a sudden increase in the number of users and model calls. Cloud hosting such as Function Compute has the advantages of free operation and maintenance, automatic elasticity, and automatic disaster recovery, ensuring service experience while reducing costs and increasing efficiency.

• Traditional hosting is inefficient

From the description of the MCP architecture, we can see that MCP Server, as the middle layer between large AI models and enterprise services, is inefficient to deploy by purchasing traditional cloud resources. Its code is usually relatively lightweight, and developers need to deploy quickly. Quick testing may only require one NPX command. "MCP Servers: Lightweight programs that each expose specific capabilities through the standardized Model Context Protocol".

• Uncertain business scale

As a replacement for the original Function Calling, the scale of tool call requests has significant uncertainty. Traditional cloud resource hosting requires long-term resource holding, and resource supply cannot be flexibly and dynamically adapted according to business traffic.

• The custom expansion process is complicated

As the middle layer of AI and enterprise service capabilities, MCP Server has logic covering simple routing to complex computing. As business scenarios become richer, it will become more complex. When choosing cloud hosting, you must also consider the efficiency of subsequent business development and maintenance. It is necessary to require more flexible customization capabilities at the development level to achieve rapid changes, rapid launch, and flexible version and traffic management.

• Data access network configuration is complex

Traditional MCP Servers rely on local deployment to achieve data security. With the popularization of cloud deployment, cloud-based MCP Servers not only need to be able to access enterprise private data in real time and securely, but also need to adapt to complex business environments and communicate between the Internet and Intranet networks. This requires the ability to quickly connect Internet public services and enterprise cloud VPCs to provide a secure and flexible execution environment.

The community actively promotes the evolution of the MCP protocol and promotes the Steamable HTTP transport technology to replace the original HTTP+SSE communication method. The original MCP transmission method is like you have to stay online when talking to customer service (SSE requires a long connection), while the new method is more like you can send a message at any time and wait for a reply (ordinary HTTP request, but can be streamed). This form is more in line with the stateless mode of Serverless computing power. The evolution of the protocol layer will be more conducive to the value amplification of Serverless computing power on the cloud. As the complexity of AI models and the scale of data continue to grow, the combination of Serverless and MCP Server will become a trend.

https://github.com/modelcontextprotocol/specification/pull/206

As a typical representative of Serverless computing power on the cloud, Function Compute directly addresses the core pain points of MCP Server hosting on the cloud with its product capabilities such as development efficiency, pay-as-you-go, and extreme elasticity, providing enterprises with efficient, flexible, and business-scale hosting capabilities for MCP Server.

• Pay on demand to avoid wasting resources
  ：Serverless is charged based on actual computing resource consumption rather than fixed server rental fees, and is particularly suitable for fluctuating loads commonly seen in AI training and reasoning tasks.
• Eliminate idle costs
  ：AI model training usually requires bursty computing power. Serverless can automatically allocate resources to avoid the problem of server idling caused by reserving resources in the traditional model.

• Dynamic resource allocation
  ：Hosting MCP Server on Function Compute, based on the Serverless architecture, can respond to AI task requirements in real time, automatically expand CPU/GPU, and ensure high concurrent processing capabilities of computing power.
• Multi-model collaboration support
  : Supports multiple AI projects running in parallel, dynamically scheduling resources according to priority, and improving overall computing power utilization.

• Serverless management
  ：Developers do not need to worry about server configuration, patch updates or cluster management, and can focus on algorithm optimization and iteration of MCP Server's internal logic and tool richness.
• Out-of-the-box toolchain
  ：Function Compute provides a complete tool chain capability and enables rapid local deployment based on the open source Serverless Devs open source tools.

• Currently, Function Compute provides single-instance multi-concurrency capabilities and expands the adaptation to the existing SSE protocol. Based on the MCP Proxy solution provided by the community, the existing local MCP Server can be quickly hosted on the cloud, facilitating the testing and development of the business platform.
• Provides a reference implementation of MCP protocol adaptation based on WebSocket, supports single-instance single concurrency and single-instance multi-concurrency capabilities, and improves protocol adaptation and scenario adaptation; at the same time, the team keeps up with the community Streamable HTTP solution, so stay tuned!

Relying on the Serverless application development platform  CAP , we can quickly implement one-click hosting of the open source MCP Server. If you need to add navigation services to the AI ​​Agent you build, you may need the MCP Server provided by the Amap community. Next, we will use the open source project amap-maps-mcp-server as an example to demonstrate how to deploy the MCP Server to Function Compute FC with one click.

Click  https://cap.console.aliyun.com/create-project?template=start-mcp-amap-maps  to enter the CAP console. Fill in the Token applied for from Amap developers (application completed immediately), which can be applied for here (https://lbs.amap.com/api/webservice/create-project-and-key) .

After successful deployment, you can get the test URL through the trigger page to test the current MCP Server. If you want to use the deployed MCP Server for production, it is recommended to use a custom domain name instead of the test URL.

Test step 1: Run the command in the local terminal:  npx @modelcontextprotocol/inspector

Test step 2: Open the locally provided test address " http://localhost:5173/#tools " in the browser for testing, fill in the URL obtained above in the URL form, add the /sse suffix and fill in the URL form, click Connect to see the Tools list provided by the open source MCP Server, and click the top Tool for interactive verification.