Demo de MCP Server Customizado no OpenShift
osa-ora/ocp-ai-custom-mcp-demo
Deploy de MCP Server customizado em OpenShift com HR DB
Ver no GitHubDemonstração prática de como deployar um servidor MCP customizado (HR) sobre OpenShift, incluindo configuração de banco de dados PostgreSQL, ConfigMaps, Secrets e integração com OpenShift AI. Fornece guia passo-a-passo completo desde a criação do projeto até testes e auto-descoberta.
Instalação
npx convez-ai osa-ora/ocp-ai-custom-mcp-demo1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Aprender padrões de deployment de MCP servers em Kubernetes/OpenShift
- Integrar servidores MCP customizados com OpenShift AI
- Provisionar aplicações com S2I (Source-to-Image) e configurações dinâmicas
Veredito

Projeto técnico robusto com documentação clara para deployar MCP servers em ambientes Enterprise OpenShift. Ideal para times DevOps e arquitetos buscando integrar AI com infraestrutura existente, mas requer conhecimento prévio de Kubernetes/OpenShift.
README
A tradução para o português deste README ainda está na fila.
OCP AI Custom MCP Server Demo
Demo to show how to deploy our custom HR MCP Server over OpenShift.
This is how it looks when this MCP Server deployed into OpenShift:
Steps to Run the Demo on OpenShift Environment
1) Create Project and Configurations
After login to OpenShift cluster, follow these steps:
# Create new project
oc new-project hr-mcp
# Create ConfigMap
oc create configmap hr-mcp-config \
--from-literal=DATABASE_USER=postgres \
--from-literal=DATABASE_HOST=postgresql.hr-mcp.svc.cluster.local \
--from-literal=DATABASE_PORT=5432 \
--from-literal=DATABASE_NAME=hrdb \
--from-literal=MCP_SCHEME=http \
--from-literal=MCP_HOST=0.0.0.0 \
--from-literal=MCP_PORT=8080 \
--from-literal=MCP_PATH=/mcp \
--from-literal=MCP_TRANSPORT=http \
--from-literal=DEBUG=true \
-n hr-mcp
# Create Secret
oc create secret generic hr-mcp-secret \
--from-literal=DATABASE_USER=postgres \
--from-literal=DATABASE_PASSWORD=postgres \
-n hr-mcp
2) Deploy Our HR DB and initialize it
Execute the following command (or you can provision it from the OCP GUI):
oc new-app postgresql-persistent \
--param=POSTGRESQL_USER=postgres \
--param=POSTGRESQL_PASSWORD=postgres \
--param=POSTGRESQL_DATABASE=hrdb \
--name=postgresql -n hr-mcp
Login to the DB Pod and run the following commands:
psql -U postgres
CREATE DATABASE hrdb;
\q
Then install the DB schema and test data from inside the Pod:
curl -s https://raw.githubusercontent.com/osa-ora/ocp-ai-custom-mcp-demo/refs/heads/main/db_scripts/schema.sql \
| psql -U postgres -d hrdb
curl -s https://raw.githubusercontent.com/osa-ora/ocp-ai-custom-mcp-demo/refs/heads/main/db_scripts/load_sample_data.sql \
| psql -U postgres -d hrdb
3) Provision the mcp_server using S2I
Provision it and attach the configMap and Secret
oc new-app python:3.12-minimal-ubi10~https://github.com/osa-ora/ocp-ai-custom-mcp-demo --name=hr-mcp-server -n hr-mcp
oc set env deployment/hr-mcp-server --from=configmap/hr-mcp-config
oc set env deployment/hr-mcp-server --from=secret/hr-mcp-secret
oc rollout restart deployment/hr-mcp-server
if started successfully, check the logs to confirm that no issues and end points are correctly mapped:
4) Testing the MCP Server:
If you want to test it remotely then expose a route:
oc expose svc/hr-mcp-server -n hr-mcp
if you need to test it from inside OpenShift using the playground, then go to project redhat-ods-applications and edit the ConfigMap: gen-ai-aa-mcp-servers by adding the following section (Service endpoint/mcp) :
kind: ConfigMap
apiVersion: v1
metadata:
name: gen-ai-aa-mcp-servers
namespace: redhat-ods-applications
....
....
data:
HR-MCP-Server: |
{
"url": "http://hr-mcp-server.hr-mcp.svc.cluster.local:8080/mcp",
"description": "An MCP server for interacting with our HR DB tools."
}
Or you can create MCPServer to auto-discovered using the OpenShift AI MCP server section
Something like this:
oc apply -f - <<EOF
apiVersion: mcp.opendatahub.io/v1alpha1
kind: MCPServer
metadata:
name: my-hr-mcp-server
namespace: hr-mcp
spec:
image: image-registry.openshift-image-registry.svc:5000/test/hr-mcp-server@sha256:549f50c109f4c6d040e948edf3169fc811be7559ec171f66062fa06f0ad60929
transportType: sse
port: 8000
envFrom:
- configMapRef:
name: hr-mcp-config
- secretRef:
name: hr-mcp-secret
EOF
But i didn't tested it, as i don't have the MCP Operstor enabled in my environment.
Now go to OpenShift UI, check the MCP Servers section:
Then create a playground for the MCP server and one of the existing models, and enjoy chatting to it:
Try to modify the prompt to give the LLM guidelines how to use the HR MCP server, for example: "if you don't have the employee_code try to get it first before calling other functions."
To connect to this KCP server remotely from outside OpenShift, for example: ChatBox application, get the route URL or execute the following command.
oc get route hr-mcp-server -n hr-mcp -o jsonpath='{.spec.host}{"\n"}'
Add the http prefix and the path configured /mcp at the end, and configure it in ChatBox or any other applicaton as following:
And enjoy chatting with our HR MCP Server...
Example requests:
- leave balance for Osama Oransa?
- basic profile for Sara Ali
- Show my full profile for EMP001?
- policy for remote work?
- leave requests for Osama Oransa.
- basic profile for EMP002