Deploy AI Model Context Protocol (MCP) on Alibaba Cloud Serverless

What is MCP and why it matters

In November 2024 Anthropic released the Model Context Protocol (MCP), an open standard that standardises the two‑way interaction between large language models (LLMs) and external services, similar to how USB unifies device connections.

Current ecosystem challenges

Localization bottleneck : most MCP servers use STDIO without OAuth2.1 authentication, leading to debugging difficulty and security risks.

Elasticity issues : traffic is bursty, causing resource waste with traditional VM deployments.

Development friction : moving from local debugging to cloud requires extensive refactoring of auth, env variables, logging, etc.

Serverless as the optimal hosting solution

Serverless provides event‑driven, millisecond‑scale elasticity, pay‑per‑use, and sandboxed runtime, perfectly addressing performance, cost and security needs of MCP services. Alibaba Cloud Function Compute supports Node.js and Python MCP servers out of the box.

Quick deployment methods

1. One‑click template deployment

Choose an MCP template in the FunctionAI console and deploy.

2. Custom MCP service

Create a blank project, add an MCP service, configure, and deploy.

3. Local deployment with ServerlessDevs

Install the tool and run the commands:

npm install @serverless-devs/s -g

s init start-fcai-mcp-nodejs

s deploy

Authentication

Only SSE protocol supports OAuth2.1; FunctionAI can generate a Bearer token to protect MCP services.

Variable management

Secrets such as API keys can be stored as service variables and injected as environment variables or startup parameters.

Continuous deployment via code repository

Bind a Git repository (GitHub, Gitee, GitLab, etc.) to enable automatic deployment on push.

Fast mode for latency‑sensitive workloads

Pre‑reserve instance snapshots to reduce cold‑start latency; only active requests incur vCPU cost.

Summary and outlook

FunctionAI now offers end‑to‑end MCP development: template, custom, and ServerlessDevs deployments; automatic STDIO/SSE hosting; console and local debugging; variable management, authentication, CI/CD, monitoring, and fast mode. Future work includes tighter integration with agents and session affinity.