mcp -operator

The MCP Server Operator is a Kubernetes/OpenShift operator that manages the lifecycle of MCP (Model Context Protocol) servers through a Custom Resource Definition (CRD) named MCPServer. Users create MCPServer resources to specify which container image should run as an MCP server and what runtime arguments should be passed to that container. The operator handles deployment, updates, and cleanup of MCP server instances on the cluster, making it easier to run and scale MCP services within OpenShift clusters. You can leverage the CRD to deploy multiple MCP servers with different configurations and images as needed by your data processing pipelines or inference services.

To use it, install the operator on your cluster and then create MCPServer objects that describe the desired server image and command-line arguments. The MCPServer spec supports fields like image, args, and command to customize how the server container is started. Once the MCPServer resource exists, the operator reconciles it into a running deployment, ensuring the specified image is used and the provided arguments are applied. If you need to update the server, modify the MCPServer resource; the operator will roll out the changes. If you no longer need a server, delete the MCPServer object and the operator will clean up the associated resources.

Prerequisites

• An OpenShift cluster (ROSA, OSD) or a compatible Kubernetes cluster
• The OpenShift CLI tool: oc
• A working Go development environment (for local development) and Kubebuilder (optional for development), if you plan to build from source
• Access to install CRDs on the cluster

Installation steps

1. Clone the repository and switch to the operator folder

git clone https://github.com/opendatahub-io/mcp-server-operator.git
cd mcp-server-operator

2. Log in to your OpenShift cluster (or your Kubernetes cluster if applicable)

oc login --token=<your user token> --server=<your openshift cluster server>

3. Set the image for deployment (adjust as needed)

export IMG=<your-registry>/<username>/mcp-server-operator:<tag>

4. Build and install CRDs

make install

5. Run the operator locally (for development/testing)

make run

6. Deploy the operator on a cluster

make deploy IMG=$IMG

Notes

• If you prefer a direct containerized run, you can adapt the docker-based command once you have built or pulled the operator image.
• Ensure you have the necessary permissions to install CRDs and create deployments in the target namespace (commonly the operator namespace such as mcp-server-operator-system).

Tips and common issues:

• Ensure your cluster has the necessary RBAC permissions for the operator to manage MCPServer CRDs and deployments.
• If you see CRD not found errors, verify that make install successfully applied the MCPServer CRD definitions.
• When upgrading the operator, review changes to the MCPServer CRD schema to avoid breaking existing resources.
• Use a dedicated namespace (e.g., mcp-server-operator-system) for running the operator and MCPServer resources.
• The MCPServer spec supports image, args, and command. Use image to pin the container image and args/command to customize startup behavior.
• For OpenShift-specific environments, ensure that your cluster has the required security contexts and permissions for running containerized MCP servers.