kong

Why we included this project

Kong has long been a default pick for teams that want one gateway in front of many microservices, handling routing, authentication, rate limiting, and observability in a single place. What makes it interesting now is that the same gateway also speaks LLM and MCP: you can proxy several model providers, enforce semantic security, and watch agent traffic without standing up a second layer. It runs natively on Kubernetes through the official ingress controller, so it fits into an existing cloud-native setup rather than demanding its own control plane. The plugin system is where most of the value lives, letting you add community or custom behavior without forking the core. If your API estate is already growing toward AI workloads, Kong is a solid thing to test.