Insights

The MCP server lets you connect lightweight clients to Red Hat Insights data sources and actions, enabling automated interactions with services like advisor, inventory, vulnerability, planning, remediations, and image building. You install it once, configure your client connections, and then issue MCP requests from your tooling to query data, run analyses, or trigger actions across Insights services.

How to use

You connect an MCP client to the Insights MCP server to perform data access and operations across Insights toolsets. Start by choosing your preferred transport: STDIO for local development or HTTP for networked integration.

• STDIO (local) lets you run the MCP server in a container and interact with it through standard input/output. This is ideal for development, debugging, and IDE integrations that spawn a local process.

• HTTP exposes the MCP API over a network URL, suitable for desktop apps, CI workflows, or services that can reach the server. You’ll provide authentication headers or tokens as described in your client setup.

How to install

Prerequisites: you need a container runtime and access to run containers. The following instructions assume Podman is installed; Docker works with equivalent command syntax where noted.

STEP 1: Install Podman (or verify Docker availability) on your system.

STEP 2: Prepare credentials for Insights APIs (if you plan to use the Insights API-enabled features). You will typically obtain a Client ID and Client Secret from your organization’s service account management portal. You will use these values in your MCP client configuration.

STEP 3: Start an MCP server instance in STDIO mode (local development). Use the following configuration with Podman to run the MCP container and expose the necessary environment variables.

Configuration and usage notes

Credentials and permissions: The MCP server authenticates to Red Hat Insights APIs using a service account. Your service account should have the appropriate roles for the toolsets you intend to use, such as RHEL Advisor viewer, Inventory Hosts viewer, Vulnerability viewer, and Remediations user, depending on your needs.

Security reminder: If you run the MCP server locally, ensure it is not exposed publicly. When working across machines or environments, be mindful of how you handle INSIGHTS_CLIENT_ID and INSIGHTS_CLIENT_SECRET. If unsure, disable or remove these credentials after you finish your session.

Client integrations: You can configure several clients to connect to the MCP server, including VSCode, Cursor, Gemini, Claude Desktop, and other tooling. Each client has its own setup flow and may require you to provide the MCP URL and authorization headers or credentials.

Troubleshooting and tips

If you encounter connectivity issues, verify that the MCP server is reachable at the configured URL or socket, and ensure that the INSIGHTS_CLIENT_ID and INSIGHTS_CLIENT_SECRET (if used) are correct. Check container logs for startup errors and verify that the required environment variables are passed to the container.

For development workflows, start with STDIO mode to isolate client-side integration before attempting HTTP-based transport. This minimizes network-related variables during debugging.