Configuring Content Cortex AI Services to support external MCP clients
Configure the Content Cortex AI Services Core MCP Server to support external Model Context Protocol (MCP) clients by registering an OAuth client in IBM Identity Management and updating the CCXAIServices operator configuration.
Before you begin
Before you begin, make sure that you have the following prerequisites:
- Admin access to Cloud Pak for Data
- OpenShift CLI (oc) installed and logged in to your cluster
- Access to the namespace where the MCP server is deployed
Set the following environment variables for your environment:
# Your Cloud Pak for Data URL
export CPD_URL="https://<your-cp4d-hostname>"
# Your MCP server public URL
export MCP_URL="https://<your-mcp-server-hostname>"
# Your OpenShift namespace where MCP server is deployed
export NAMESPACE="<your-namespace>"
Example:
export CPD_URL="https://cpd-adp.apps.rosier.cp.fyre.ibm.com"
export MCP_URL="https://core-mcp-adp.apps.rosier.cp.fyre.ibm.com"
export NAMESPACE="adp"
About this task
To configure Content Cortex AI Services to support external MCP clients, complete the following steps.
Note: The
redirect_uris parameter must match the redirect URIs of your MCP client. For watsonx Orchestrate, see MCP servers.Procedure
Results
After you complete the steps, the Content Cortex AI Services Core MCP Server is configured to support external MCP clients with OAuth authentication.
What to do next
Verify that all resources were created correctly:
- Secrets exist:
mcp-oauth-client-secretandibm-core-mcp-secret(Redis only) - ConfigMaps exist:
ibm-ai-services-integration-configandcore-mcp-override-custom-configmap - CCXAIServices custom resource
ai-servicesis updated - MCP server pod is running with correct environment variables (
AUTH_MODE=dualandMCP_SERVER_BASE_URLset correctly)