Skip to main content

Set up the IO MCP in Global View

Setting up MCP for Global View on-prem is required because on-prem deployments do not pick up IO integration automatically from MCP connection details alone. You must configure the IO_MCP_ENDPOINT in the Global View deployment values and apply the change through redeployment or a service/pod restart so Global View Copilot can load the endpoint, authenticate successfully, and retrieve IO data.

Note

This setup applies only to Global View on-prem deployments. Global View SaaS environments connect to Infrastructure Observability automatically and do not require these steps.

Prerequisites

Before you configure the MCP endpoint, make sure the following requirements are met:

  • Ensure that Global View is already deployed and running. You can add the IO MCP integration only after the Global View deployment is in place.

  • Ensure that the IO server has a valid SSL certificate. The integration does not work if the server certificate is invalid or untrusted.

  • Ensure that you have the IO MCP endpoint URL and a valid Bearer token.

Caution

A valid SSL certificate on the IO server is a hard prerequisite for Global View to IO MCP integration. If the certificate is invalid, the integration fails even when the endpoint and authentication headers are correct.

Configure the IO MCP endpoint in Global View

To register your IO instance with Global View, add the endpoint to the deployment configuration:

  1. Open the globalview-values.yaml file for your Global View deployment.

  2. Locate the cp-configs section.

  3. Add or update the IO_MCP_ENDPOINT setting with the endpoint configuration for your tenant and instance.

    cp-configs:
  IO_MCP_ENDPOINT: '{"<tenant-id>": {"<instance-name>": {"url": "https://<io-server>/api/sdk/mcp/", "headers": {"Authorization": "Bearer <your-io-token>"}}}}'

    Replace the placeholder values as follows:

    Table 10.

    Placeholder

    Description

    <tenant-id>

    Your tenant identifier.

    <instance-name>

    A name for the target IO instance.

    <io-server>

    The hostname of the IO server.

    <your-io-token>

    A valid API bearer token.



Apply configuration

After you update globalview-values.yaml, apply the change to the deployment.

  • If you are configuring MCP as part of a planned deployment update, redeploy the Infra services so the new configuration is picked up.

  • If Global View is already deployed and the MCP endpoint is being added afterward, delete or restart the virtana-ai pod so it reloads the updated configuration.

To restart the pod, delete it. Kubernetes recreates the pod automatically. Run the following command:

kubectl delete pod <virtana-ai-pod-name> -n <namespace>

Note

If the MCP endpoint is configured after an initial Global View deployment or upgrade, restarting the virtana-ai pod is required so the updated endpoint configuration becomes active.

Verify the integration

To confirm the integration works, query the Copilot for data that comes from IO:

  1. Log in to the Virtana Platform and open Global View.

  2. Open the Copilot.

  3. Submit a prompt that requires IO data.

  4. Confirm that the response includes data from IO.

    If it does, Global View Copilot reached the MCP endpoint and retrieved the data successfully.

In this section: