MCP Gateway on Kubernetes Using Agent Gateway

Source: MCP Gateway on Kubernetes using Agent Gateway — That DevOps Guy (https://www.youtube.com/watch?v=_EfxYSpo2mo) · published 2026-06-04
Speaker(s): Not stated
Relevant to: Agentic OS, BexarByte, Cross-business

TL;DR

This video explains how to deploy and configure Agent Gateway on Kubernetes to route traffic to MCP servers, which are tools used by large language models (LLMs) to interact with systems like Kubernetes. It covers the setup of MCP servers, the use of Gateway API for routing, and the concept of MCP federation to unify multiple MCP endpoints under a single URL. This is particularly relevant for AI-native operations that need secure, scalable, and declarative routing for AI agents and MCP tools.

Key ideas

• [00:02] Model Context Protocol (MCP) allows LLMs to interact with systems like Kubernetes by using predefined tools.
• [01:12] MCP servers provide deterministic access to systems, avoiding the pitfalls of LLMs guessing API calls.
• [02:08] MCP servers can be deployed as Kubernetes pods and secured using standard web security practices.
• [02:50] Agent Gateway routes traffic to both LLMs and MCP servers using Kubernetes Gateway API constructs.
• [04:43] A Kubernetes cluster is created using kind for local testing.
• [05:04] Gateway API CRDs are installed to enable advanced traffic routing.
• [06:13] Agent Gateway is installed using Helm charts for both CRDs and the control plane.
• [08:42] Gateways are defined using YAML files to route traffic to MCP servers.
• [11:31] MCP servers are deployed as Kubernetes services with read-only service accounts.
• [13:10] Agent Gateway backends are defined to route traffic to specific MCP servers.
• [14:02] HTTP routes are used to expose MCP servers via URLs, with URL rewriting for client-facing endpoints.
• [20:21] MCP federation allows multiple MCP servers to be accessed under a single endpoint, improving usability for AI agents.

Tools, services & specific callouts

Tool / service	What it is	How it's used in the video	Use-case for my ventures
Agent Gateway	A routing proxy for AI workloads, including LLMs and MCP servers.	Used to route traffic to MCP servers via Kubernetes Gateway API.	Enables secure, scalable routing for AI agents and MCP tools in Agentic OS.
Kubernetes Gateway API	A specification for managing traffic in Kubernetes.	Used to define gateways, HTTP routes, and backend services.	Enables declarative, secure traffic management for AI workloads in Agentic OS and BexarByte.
kind	A tool for creating Kubernetes clusters locally.	Used to create a local Kubernetes cluster for testing.	Useful for local development and testing of AI-native infrastructure in Agentic OS.
Helm	A package manager for Kubernetes.	Used to install Agent Gateway and its CRDs.	Streamlines deployment of AI-native routing infrastructure in Agentic OS.
MCP server	A tool that allows LLMs to interact with systems like Kubernetes.	Deployed as a Kubernetes service with a read-only service account.	Enables secure, deterministic access to systems from AI agents in Agentic OS.
GitHub MCP server	A tool for interacting with GitHub via MCP.	Deployed as a Kubernetes service with a service account.	Useful for AI agents needing to interact with GitHub in BexarByte or Agentic OS.

How AI / automation is used

• MCP servers are used to provide deterministic access to systems like Kubernetes, avoiding the risk of LLMs making incorrect API calls.
• Agent Gateway routes traffic to both LLMs and MCP servers using Kubernetes Gateway API, enabling a unified interface for AI agents.
• HTTP routes with URL rewriting allow clients to access MCP servers via a single, consistent endpoint.
• MCP federation allows multiple MCP servers to be accessed under a single endpoint, improving usability for AI agents.

SEO / GEO / marketing / growth

• None in this video.

Infrastructure & hardware

Container / protocol	Description	Use-case for my ventures
kind	Creates a local Kubernetes cluster for testing.	Useful for local development and testing of AI-native infrastructure in Agentic OS.
Kubernetes Gateway API	Manages traffic routing in Kubernetes.	Enables secure, scalable traffic management for AI workloads in Agentic OS.
Helm	Installs Agent Gateway and its CRDs.	Streamlines deployment of AI-native routing infrastructure in Agentic OS.
Service accounts with read-only permissions	Used to secure access to Kubernetes and GitHub.	Ensures secure access to systems from AI agents in Agentic OS.

Notable quotes

"MCP servers provide a set of tools that large language models can use to access various systems." — [00:10]
"MCP exists to make tool execution more deterministic as the model does not have to guess." — [01:30]
"Agent Gateway allows us to route traffic to large language models, route traffic to agents and MCP servers." — [02:31]
"The fundamentals of web security do not go away just because we live in an AI world." — [02:16]
"MCP federation allows us to federate all of our MCP tools under a single endpoint." — [20:16]

Actionable takeaways

• [Agentic OS] Deploy MCP servers as Kubernetes services with read-only service accounts to ensure secure access.
• [Agentic OS] Use Agent Gateway with Kubernetes Gateway API to route traffic to MCP servers and LLMs.
• [BexarByte] Implement MCP federation to allow AI agents to access multiple systems under a single endpoint.
• [Cross-business] Use Helm to streamline the deployment of AI-native routing infrastructure.

Open questions / to verify

• The video mentions installing the "1.4 experimental version" of Gateway API CRDs, but does not provide a source for this version.
• The GitHub repository for the source code is mentioned but not linked explicitly in the transcript.

Filing metadata

• Suggested title: "MCP Gateway on Kubernetes Using Agent Gateway"
• Primary venture: Agentic OS
• Secondary ventures: BexarByte, Cross-business
• Type: Tooling/How-to
• Keyword tags: MCP, Kubernetes, Agent Gateway, Gateway API, AI routing, DevOps, AI-native
• One-line index hook: This article explains how to deploy and configure Agent Gateway on Kubernetes to route traffic to MCP servers, enabling secure and scalable AI agent interactions with systems like Kubernetes.

---

AI-assisted summary of a YouTube video — source. Generated by a local model and human-reviewed; verify specifics against the original before relying on them.