AI Observability Summarizer

What it does

This MCP server connects to OpenShift and OpenShift AI observability data (Prometheus, Thanos, DCGM) to provide AI-driven summaries of model performance, GPU utilization, and cluster health. It eliminates the need to manually stitch together dashboards by providing a natural-language interface to observability signals.

Tools

• analyze_metrics: Generates summaries of specific Prometheus metric trends.
• generate_observability_report: Creates structured HTML/PDF/Markdown reports of cluster state.
• chat_with_prometheus: Translates natural language questions into PromQL queries and explains results.
• get_gpu_metrics: Fetches hardware accelerator health and utilization data.

Installation

Exact claude_desktop_config.json snippet depends on the deployment mode (Helm/Operator). Typically configured as an HTTP/SSE server:

{
  "mcpServers": {
    "ai-observability": {
      "url": "http://your-openshift-route/mcp"
    }
  }
}

Supported hosts

Confirmed integration with OpenShift Console and general MCP clients.