Apache APISIX logo

Apache APISIX

Dynamic, high-performance API gateway for modern cloud and AI workloads

Apache APISIX delivers real-time traffic management, load balancing, security, and observability across HTTP, gRPC, TCP/UDP, and AI services, with hot-reload plugins and cloud-native deployment options.

Apache APISIX banner

Overview

Overview

Apache APISIX is a real-time, high‑performance API gateway that handles north‑south and east‑west traffic for microservices, AI models, and traditional web APIs. It supports a wide range of protocols—including HTTP/HTTPS, gRPC, WebSocket, TCP/UDP, MQTT, Dubbo, and HTTP/3 with QUIC—allowing heterogeneous services to be exposed through a single entry point.

Capabilities & Deployment

The gateway offers dynamic routing, canary releases, A/B testing, circuit breaking, and token‑based rate limiting, all configurable via hot‑reload without service interruption. Security is reinforced with JWT, OAuth, RBAC, IP/Referer whitelists, and CSRF protection. Observability integrates natively with Prometheus, Zipkin, SkyWalking, and multiple external loggers. APISIX runs statelessly, scaling horizontally across bare metal or Kubernetes clusters, and relies on etcd or compatible service‑discovery backends for configuration storage.

AI Gateway Extension

Through its plugin system, APISIX can act as an AI gateway, providing LLM load balancing, retries, fallbacks, and token‑rate controls, ensuring reliable and efficient AI agent interactions.

Highlights

Hot-reload configuration and plugins without restarts
Multi-protocol support: HTTP, HTTPS, gRPC, WebSocket, TCP/UDP, MQTT, Dubbo
Advanced traffic control: canary releases, A/B testing, circuit breaking, rate limiting
Extensive security integrations: JWT, OAuth, RBAC, IP/Referer whitelists

Pros

  • High throughput (≈18k QPS per core) with sub-millisecond latency
  • Stateless nodes enable easy horizontal scaling and clustering
  • Rich plugin ecosystem, including AI gateway extensions
  • Native integration with Prometheus, Zipkin, SkyWalking for observability

Considerations

  • Configuration complexity can be steep for beginners
  • Requires external etcd or compatible service discovery
  • Advanced features may need custom plugins or Lua scripting
  • Enterprise-grade support relies on community contributions

Managed products teams compare with

When teams consider Apache APISIX, these hosted platforms usually appear on the same shortlist.

Amazon API Gateway logo

Amazon API Gateway

Fully managed service to create, publish, and secure APIs at any scale for backend access

Azure API Management logo

Azure API Management

Fully managed multicloud API management service for publishing, securing, and monitoring APIs across environments

Google Apigee API Management logo

Google Apigee API Management

API management platform to publish, secure, and analyze APIs

Looking for a hosted option? These are the services engineering teams benchmark against before choosing open source.

Fit guide

Great for

  • Enterprises needing flexible, cloud-agnostic API management
  • Teams deploying AI/LLM services that require token-based rate limiting
  • Organizations adopting microservices with diverse protocols
  • Operators who value hot-reload and zero-downtime deployments

Not ideal when

  • Small projects that prefer a simple, single-file proxy
  • Environments without etcd or compatible key-value store
  • Teams lacking Lua or plugin development expertise
  • Use cases demanding out-of-the-box commercial SLA

How teams use it

AI model serving gateway

Balances LLM requests, applies token-based throttling, and retries failed calls, ensuring consistent latency for AI applications.

Multi-protocol microservice ingress

Routes HTTP, gRPC, and MQTT traffic to appropriate services with dynamic load balancing and health checks.

Canary deployment for new API version

Gradually shifts a percentage of traffic to the new version while monitoring error rates, enabling safe rollouts.

Zero-downtime configuration updates

Updates routing rules and plugins on the fly without restarting services, maintaining uninterrupted API availability.

Tech snapshot

Lua80%
Shell11%
Go3%
TypeScript2%
Perl1%
Python1%

Tags

kubernetesapisixcloud-nativeservice-meshiotapi-managementnginxloadbalancingapisluajitkubernetes-ingress-controllerapigatewaymicroservicesapi-gatewaydevopsserverlessapireverse-proxydockerkubernetes-ingress

Frequently asked questions

How does APISIX achieve hot-reload without downtime?

Configuration and plugins are stored in etcd; APISIX watches for changes and applies them in-process, so updates take effect instantly without restarting the gateway.

What protocols are natively supported?

HTTP/HTTPS, gRPC, gRPC-Web, WebSocket, TCP, UDP, MQTT (3.1.*, 5.0), Dubbo, and HTTP/3 with QUIC.

Can APISIX run in Kubernetes?

Yes, it provides a native ingress controller and Helm charts for deployment on Kubernetes clusters.

Which observability tools integrate out of the box?

Prometheus for metrics, Zipkin and SkyWalking for tracing, and various external loggers such as Elasticsearch, ClickHouse, and Datadog.

What authentication methods are available?

Key-auth, JWT, basic-auth, OAuth via external IdPs (Keycloak, Auth0, Okta), RBAC plugins like casbin, and IP/Referer whitelists.

Project at a glance

Active
Stars
16,101
Watchers
16,101
Forks
2,773
LicenseApache-2.0
Repo age6 years old
Last commit2 days ago
Primary languageLua

Last synced 4 hours ago