higress: Open Source API Development & Testing

What is higress?

Higress is a cloud-native, AI-native API gateway built on Istio and Envoy that targets platform engineering and AI infrastructure teams who need to route LLM API traffic and Model Context Protocol (MCP) servers through a single extensible, self-hosted gateway.

What it is

Higress is an API gateway that lives in the Kubernetes and cloud-native ecosystem, built on top of Istio and Envoy and distributed as a Go project under the Apache-2.0 licence. Its distinguishing trait is that it is AI-native: alongside conventional gateway duties it handles LLM API traffic, supports mainstream model providers both domestic and international, and can host MCP servers through its plugin mechanism so that AI agents can call tools and services. Extension happens through Wasm plugins written in Go, Rust, or JavaScript, and the project ships dozens of ready-to-use general-purpose plugins plus an out-of-the-box console.

The concrete problems it was built to solve are specific. Higress originated at Alibaba to address long-connection disruption during gateway reloads and to improve gRPC and Dubbo load balancing. It also solves the fragmentation problem that appears once a platform team is running both model APIs and tool APIs: Higress provides unified management for both LLM API and MCP API, with the openapi-to-mcp tool available to convert OpenAPI specifications into remote MCP servers for hosting, rather than requiring a separate tool for each category. The gateway tier it occupies is the ingress and API traffic layer of a Kubernetes deployment.

Key capabilities

Wasm plugin extension model, with plugins authored in Go, Rust, or JavaScript and published through the Wasm Plugin Hub.
AI gateway support for mainstream model providers, both domestic and international, via the ai-proxy plugin and its provider directory under plugins/wasm-go/extensions/ai-proxy/provider.
MCP server hosting through the plugin mechanism, letting AI agents call tools and services.
openapi-to-mcp conversion of OpenAPI specifications into remote MCP servers.
Unified management of LLM API and MCP API traffic in one control surface.
Out-of-the-box console served on port 8001, with a public demo at demo.higress.io and a hosted MCP server platform at https://mcp.higress.ai/.
Envoy and Istio as the underlying data and control plane foundations.

Who uses it and how

Platform teams replacing reload-disruptive gateway behaviour: the original Alibaba use case was long-connection disruption during gateway reloads, plus gRPC and Dubbo load balancing.
Organizations across multiple industries, as a vendor-neutral CNCF project; public adopters and their use cases are listed in ADOPTERS.md.
Individual developers running the all-in-one container locally for learning or for building simple sites, driven entirely through Docker on ports 8001, 8080, and 8443.
AI agent builders hosting remote MCP servers, using the hosted platform at mcp.higress.ai to try Higress-hosted servers before self-hosting.
Operators in regions outside China, who mirror the project images into a registry they control and set the Helm global.hub value accordingly.

Getting started

The README's Quick Start is a single docker run of the all-in-one image: higress-registry.cn-hangzhou.cr.aliyuncs.com/higress/all-in-one:latest, with configuration files written into the working directory and ports 8001, 8080, and 8443 exposed. Alternative regional registry endpoints are documented for North America (higress-registry.us-west-1.cr.aliyuncs.com) and Southeast Asia, and Helm deployments can point global.hub at a self-controlled mirror.

How it compares

No comparable API gateway or AI gateway product is named in the facts for this listing, and no set of paid products that Higress replaces is provided, so it stands alone in this registry rather than being positioned against a named commercial alternative. What can be stated is its licence and distribution model: Apache-2.0, self-hosted by default, with a hosted MCP server platform offered as a way to try the MCP hosting capability before running it yourself.

When to use it — and when not to

Choose Higress when a Kubernetes platform team wants one gateway for both LLM API traffic and MCP tool hosting, and is comfortable operating Envoy and Istio underneath it. A self-hoster must run the gateway and console containers and, outside China, mirror images from the regional registry endpoints to a registry they control. Teams that want a vendor-hosted gateway with no operational surface, or that do not need Wasm extensibility or MCP hosting, should look elsewhere — and the project's 1,114 open issues indicate a large active backlog rather than a small, quiet codebase.

project readme (upstream, from github) — read inline

AI Gateway

AI Native API Gateway

English | 中文 | 日本語

What is Higress?

Higress is a cloud-native API gateway based on Istio and Envoy, which can be extended with Wasm plugins written in Go/Rust/JS. It provides dozens of ready-to-use general-purpose plugins and an out-of-the-box console (try the demo here).

Core Use Cases

Higress's AI gateway capabilities support all mainstream model providers both domestic and international. It also supports hosting MCP (Model Context Protocol) Servers through its plugin mechanism, enabling AI Agents to easily call various tools and services. With the openapi-to-mcp tool, you can quickly convert OpenAPI specifications into remote MCP servers for hosting. Higress provides unified management for both LLM API and MCP API.

🌟 Try it now at https://mcp.higress.ai/ to experience Higress-hosted Remote MCP Servers firsthand:

Higress MCP Server Platform

Production Adoption

Higress originated at Alibaba to address long-connection disruption during gateway reloads and improve gRPC/Dubbo load balancing. It is now developed as a vendor-neutral CNCF project and is used by organizations across multiple industries. Public adopters and their use cases are listed in ADOPTERS.md.

Quick Start

Higress can be started with just Docker, making it convenient for individual developers to set up locally for learning or for building simple sites:

# Create a working directory
mkdir higress; cd higress
# Start higress, configuration files will be written to the working directory
docker run -d --rm --name higress-ai -v ${PWD}:/data \
        -p 8001:8001 -p 8080:8080 -p 8443:8443  \
        higress-registry.cn-hangzhou.cr.aliyuncs.com/higress/all-in-one:latest

Port descriptions:

Port 8001: Higress UI console entry
Port 8080: Gateway HTTP protocol entry
Port 8443: Gateway HTTPS protocol entry

Higress publishes project images through dedicated regional registry endpoints. Operators may mirror the images to a registry they control and configure the Helm global.hub value accordingly.

If you experience a timeout when pulling image from higress-registry.cn-hangzhou.cr.aliyuncs.com, you can try replacing it with the following docker registry mirror source:

North America: higress-registry.us-west-1.cr.aliyuncs.com

Southeast Asia: higress-registry.ap-southeast-7.cr.aliyuncs.com

For Kubernetes deployments, you can configure the global.hub parameter in Helm values to use a mirror registry closer to your region. This applies to both Higress component images and built-in Wasm plugin images:
# Example: Using North America mirror
helm install higress -n higress-system higress.io/higress --set global.hub=higress-registry.us-west-1.cr.aliyuncs.com --create-namespace
Available mirror registries:

China (Hangzhou): higress-registry.cn-hangzhou.cr.aliyuncs.com (default)

North America: higress-registry.us-west-1.cr.aliyuncs.com

Southeast Asia: higress-registry.ap-southeast-7.cr.aliyuncs.com

For other installation methods such as Helm deployment under K8s, please refer to the official Quick Start documentation.

Use Cases

MCP Server Hosting:

Higress hosts MCP Servers through its plugin mechanism, enabling AI Agents to easily call various tools and services. With the openapi-to-mcp tool, you can quickly convert OpenAPI specifications into remote MCP servers.

Key benefits of hosting MCP Servers with Higress:
- Unified authentication and authorization mechanisms
- Fine-grained rate limiting to prevent abuse
- Comprehensive audit logs for all tool calls
- Rich observability for monitoring performance
- Simplified deployment through Higress's plugin mechanism
- Dynamic updates without disruption or connection drops
  
  Learn more...
AI Gateway:

Higress connects to all LLM model providers using a unified protocol, with AI observability, multi-model load balancing, token rate limiting, and caching capabilities:

Higress is also listed as a conformant Gateway API Inference Extension implementation, supporting inference-aware routing on Kubernetes.
Kubernetes ingress controller:

Higress can function as a feature-rich ingress controller and is listed in the official Kubernetes Ingress Controllers documentation. It is compatible with many annotations of the Kubernetes NGINX Ingress Controller.

Higress supports the Gateway API standard and is listed as a conformant Gateway API implementation, enabling smooth migration from the Ingress API to the Gateway API.

Compared to ingress-nginx, the resource overhead has significantly decreased, and the speed at which route changes take effect has improved by ten times.

The following resource overhead comparison comes from sealos.

For details, you can read this article to understand how sealos migrates the monitoring of tens of thousands of ingress resources from nginx ingress to higress.
Microservice gateway:

Higress can function as a microservice gateway, which can discovery microservices from various service registries, such as Nacos, ZooKeeper, Consul, Eureka, etc.

It deeply integrates with [Dubbo](https://github.com/a

Frequently asked questions

Is higress free to use?

higress is open source under the Apache-2.0 licence. There is no licence fee and no seat count — you can self-host it or, where the project offers one, pay a vendor for a managed version instead.

What does higress do?

🤖 AI Gateway | AI Native API Gateway

What is higress written in?

higress is primarily written in Go. Its source is publicly available at https://github.com/higress-group/higress, and it has 9,397 GitHub stars.