How to Convert Alibaba HSF Services to MCP Servers Using Higress – A No‑Code Guide

Background

MCP (Model Center Protocol) standardizes resource access and multi‑agent interoperability, and its ecosystem is rapidly expanding with thousands of MCP servers. Alibaba’s internal practice shows how to join this ecosystem without modifying existing services, using the open‑source Higress AI gateway to offload MCP protocol handling.

Key Components

Higress gateway – receives MCP traffic and provides identity authentication, traffic routing, parameter mapping, and security auditing.

MCP console – platform for creating and managing MCP servers, tools, and prompts.

MCP Registry – registration center for all MCP servers, implemented by the HSF registry.

MCP Metadata Center – stores prompts, server metadata, tool metadata, and versioning, backed by the HSF configuration center.

Step‑by‑Step Migration (HSF Service → MCP Server)

Open the HSFOPS backend for the target environment and select the MCP sidebar.

Choose the HSF application you own, then specify the service name and method to be converted.

Provide a precise description for each //TODO field in the method’s input parameters (name and description).

The tool automatically generates the request‑parameter structure; you only need to add key names and descriptions.

Access the MCP endpoint using SSE: http://{MCP endpoint prefix}/{applicationName}/sse.

Challenges and Rationale

The MCP ecosystem evolves quickly; new SPEC releases appear every few months, and SDKs lag behind, requiring costly code changes and migration effort. By using a hosting approach (Higress gateway) instead of native SDK integration, developers avoid frequent breaking changes and reduce vendor lock‑in risk.

Strategic Reflections

From an AI infrastructure perspective, MCP is a useful but not mature protocol; it should be adopted cautiously, with Higress handling protocol offload while retaining existing internal protocols. This hybrid strategy enables rapid prototyping without committing to an unproven stack.

Conclusion

The described workflow lets teams convert large‑scale HSF services to MCP servers with minimal effort, accelerating entry into the MCP ecosystem and supporting AI‑native application development, while acknowledging that MCP is only the first step in a longer AI infra journey.