A Kubernetes operator that provides a declarative API to deploy, manage, and safely roll out MCP Servers, handling their full lifecycle with production-grade automation and ecosystem integrations.
Note: This project is currently in beta (). APIs and behavior may change in future releases.
A Kubernetes operator that provides a declarative API to deploy, manage, and safely roll out MCP Servers, handling their full lifecycle with production-grade automation and ecosystem integrations.
星數
50
工具
—
可用率
—
v1beta1
Documentation
Introduction - Architecture and MCPServer API overview
This installs the CRDs, the controller Deployment, and all necessary RBAC resources in the mcp-lifecycle-operator-system namespace.
Option B: Run Locally (for Development)
Clone the repository and run the controller on your local machine:
make install # Install the CRDs
make run # Run the controller locally
Keep this terminal open. The controller logs will appear here.
Option C: Build and Deploy from Source
Build and deploy the controller as a Deployment in your cluster:
# Build and push the container image for multiple platforms
make docker-buildx IMG=<your-registry>/mcp-lifecycle-operator:latest
# Deploy to cluster
make deploy IMG=<your-registry>/mcp-lifecycle-operator:latest
Note: docker-buildx builds for multiple architectures (amd64, arm64, s390x, ppc64le) and pushes automatically.
2. Create a Test MCPServer
In a new terminal, create a test MCPServer resource:
For an MCP server image in a private registry, create an image pull Secret in
the same namespace and reference it from the ContainerImage source. The
optional pullPolicy field accepts Always, IfNotPresent, or Never.
# View the MCPServer status
kubectl get mcpservers
kubectl get mcpserver test-server -o yaml
# Verify the Deployment was created
kubectl get deployment test-server
# Verify the Service was created
kubectl get service test-server
# Verify the NetworkPolicy was created
kubectl get networkpolicy test-server
# Check the pod is running
kubectl get pods -l mcp-server=test-server
Expected output from kubectl get mcpservers:
NAME READY ACCEPTED IMAGE PORT ADDRESS AGE
test-server True True quay.io/containers/kubernetes_mcp_server:latest 8080 http://test-server.default.svc.cluster.local:8080/mcp 1m
The ADDRESS column shows the cluster-internal URL that can be used by other workloads to connect to the MCP server.
The status includes conditions and the service address for easy discovery:
# Test the health endpoint
curl http://localhost:8080/healthz
# Test the MCP endpoint
curl http://localhost:8080/mcp
You should see a response from the MCP server.
5. Uninstall (Optional)
To remove the CRDs and operator:
# If you installed from the release
kubectl delete -f https://github.com/kubernetes-sigs/mcp-lifecycle-operator/releases/latest/download/install.yaml
# If you deployed from source
make undeploy
make uninstall
# Generate code and manifests
make manifests generate
# Build binary
make build
# Run unit tests
make test
# Run e2e tests (requires Kind)
make deploy-test-e2e # creates Kind cluster, builds image, deploys operator
make test-e2e # runs e2e tests against the cluster
make cleanup-test-e2e # tears down the Kind cluster
A Kubernetes operator that provides a declarative API to deploy, manage, and safely roll out MCP Servers, handling their full lifecycle with production-grade automation and ecosystem integrations.
上次檢查時間:2026-10-27。 A Kubernetes operator that provides a declarative API to deploy, manage, and safely roll out MCP Servers, handling their full lifecycle with production-grade automation and ecosystem integrations.