A guide to building long-term compounding knowledge infrastructure. See details on GitHub .

kgateway

A Kubernetes-native Envoy-powered API and AI gateway that unifies access to LLM and inference backends, supports self-hosted models, fine-grained routing, and platform-level policy management.

Overview

Kgateway is a Kubernetes-native Envoy-powered gateway that serves as both an API gateway and an AI gateway. It centralizes access to LLM and inference backends, supports self-hosted models, and provides fine-grained routing and policy controls.

Key Features

  • Multi-backend & model integrations: built-in integrations for OpenAI, Azure, Vertex, AWS Bedrock, and self-hosted inference backends.
  • Cloud-native & extensible: implemented with Gateway API and Envoy, supports Kubernetes CRDs and Helm deployments.
  • Flexible routing & policies: function-level routing, authentication, rate-limiting, header mutation, and traffic shaping.
  • Observability & operations: metrics export, logging, and OpenTelemetry support for production monitoring.

Use Cases

  • Centralize access to multiple LLM services with unified auth, rate-limiting, and auditing.
  • Deploy self-hosted inference clusters at edge or private cloud for privacy and compliance.
  • Migrate legacy services to cloud-native architectures with hybrid routing and traffic governance.

Technical Highlights

  • Built with Go and tightly integrated with the Envoy ecosystem and Kubernetes.
  • Supports MCP (Model Context Protocol) and Inference Extension for advanced model routing and context federation.

Comments

kgateway
Resource Info
Author Solo.io
Added Date 2025-09-17
Tags
OSS Deployment Dev Tools