deploying-toolhive-operator.md

Deploying ToolHive Kubernetes Operator

The ToolHive Kubernetes Operator manages MCP (Model Context Protocol) servers in Kubernetes clusters. It allows you to define MCP servers as Kubernetes resources and automates their deployment and management.

Prerequisites

• Helm installed
• Kind installed
• Optional: Task to run automated steps with a cloned copy of the ToolHive repository (git clone https://github.com/stacklok/toolhive)

TL;DR

To setup a kind cluster and/or deploy the Operator, we have created a Task so that you can do this with one command. You will need to clone this repository to run the command.

Fresh Kind Cluster with Operator Install

Run:

task kind-with-toolhive-operator

This will create the kind cluster, install an nginx ingress controller and then install the latest built ToolHive Operator image.

Existing Kind Cluster with Operator Install

Run:

# If you want to install the latest built operator image from Github (recommended)
task operator-deploy-latest

# If you want to built the operator image locally and deploy it (only recommended if you're doing development around the Operator)
task operator-deploy-local

This will install the Operator into the existing Kind cluster that your kconfig.yaml file points to.

Manual Installation

Fresh Kind Cluster with Operator Install

Follow the Kind Cluster setup guide.

Once the cluster is running, follow these steps:

1. Install the CRD:

helm upgrade -i toolhive-operator-crds oci://ghcr.io/stacklok/toolhive/toolhive-operator-crds

2. Deploy the operator:

helm upgrade -i toolhive-operator oci://ghcr.io/stacklok/toolhive/toolhive-operator -n toolhive-system --create-namespace

Existing Kind Cluster with Operator Install

1. Install the CRD:

helm upgrade -i toolhive-operator-crds oci://ghcr.io/stacklok/toolhive/toolhive-operator-crds

2. Deploy the operator:

helm upgrade -i toolhive-operator oci://ghcr.io/stacklok/toolhive/toolhive-operator -n toolhive-system --create-namespace