higress

This API gateway addresses the challenge of managing AI model APIs and long-connection disruptions during gateway reloads, particularly for gRPC and Dubbo load balancing. It builds on Istio and Envoy, adding native support for AI model providers and the ability to host MCP (Model Context Protocol) servers via Wasm plugins written in Go, Rust, or JavaScript. This makes it uniquely suited for teams integrating multiple AI services with unified management for LLM and MCP APIs. The project is moderately mature with over 8,000 stars and recent commits, but lacks notable growth momentum, suggesting a niche but active user base. Higress originated at Alibaba and now operates as a CNCF project, which may appeal to teams valuing vendor neutrality and cloud-native design. It is not the best choice for teams seeking a more established, widely adopted gateway like Kong or Ambassador that focus on broader API management without AI-specific extensions.

Choose Higress when your architecture requires integrated AI gateway features with support for multiple mainstream model providers and MCP server hosting, especially if you are already invested in Istio or Envoy. For general-purpose API gateway needs without AI focus, consider Kong or Ambassador for their larger ecosystems and more mature feature sets.

It typically fits teams building AI-native platforms that unify access to large language models and AI agents, such as internal AI service meshes or AI-enhanced microservices. Common use cases include hosting remote MCP servers converted from OpenAPI specs and managing AI model APIs alongside traditional service traffic.

Latest commit: 2026-04-21. Activity data is based on daily RepoPi snapshots of the GitHub repository.