ops-mcp-server

shaowenchen/ops-mcp-server

3.2

If you are the rightful owner of ops-mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

Ops MCP Server is a modular server for operational data querying using the Model Context Protocol (MCP), built with Go.

Tools

  1. list_events

    List all available events with optional limit and offset

  2. get_event

    Get detailed information about a specific event by ID

  3. get_metrics_status

    Get the current status of the metrics module

  4. get_logs_status

    Get the current status of the logs module

Ops MCP Server

A modular MCP (Model Context Protocol) server providing operational monitoring capabilities including events, metrics, and logs.

Features

  • Events Module: Query Kubernetes events (pods, deployments, nodes)
  • Metrics Module: Query monitoring metrics through Grafana/Prometheus
  • Logs Module: Analyze and search application logs

Metrics Module - Prometheus Integration

The metrics module supports querying monitoring data directly through Prometheus API, allowing you to execute Prometheus queries.

Available Tools

Basic Query Tools
  • query_metrics - Execute custom PromQL queries
  • query_metrics_range - Execute range queries with time intervals
  • get_metrics_status - Get metrics module status
  • get_system_overview - Get system overview metrics
Kubernetes Resource Discovery
  • get_clusters - List all available Kubernetes clusters
  • get_namespaces - List namespaces (optionally filtered by cluster)
  • get_pods - List pods (optionally filtered by cluster/namespace)
Kubernetes Resource Usage
  • get_pod_resource_usage - Get CPU and memory usage for pods
  • get_node_resource_usage - Get CPU and memory usage for nodes

Configuration

Configure Prometheus integration in configs/config.yaml:

metrics:
  enabled: true
  prometheus:
    endpoint: "http://localhost:9090/api/v1"

Environment Variables

Configure endpoint via environment variable if needed:

export METRICS_PROMETHEUS_ENDPOINT="http://localhost:9090/api/v1"

Getting Prometheus Configuration

  1. Endpoint: Your Prometheus server endpoint URL with API version (e.g., http://localhost:9090/api/v1)

Example Queries

List Clusters
{
  "tool": "get_clusters"
}
Get Pods in Specific Namespace
{
  "tool": "get_pods",
  "cluster": "my-cluster",
  "namespace": "default",
  "limit": "10"
}
Get Pod Resource Usage
{
  "tool": "get_pod_resource_usage",
  "cluster": "my-cluster",
  "namespace": "kube-system",
  "limit": "20"
}
Get Node Resource Usage
{
  "tool": "get_node_resource_usage",
  "cluster": "my-cluster",
  "limit": "10"
}
Custom PromQL Query
{
  "tool": "query_metrics",
  "query": "up{job=\"kubernetes-nodes\"}"
}

Query Examples

CPU Usage:

rate(cpu_usage_total[5m])

Memory Usage:

(1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)) * 100

HTTP Request Rate:

rate(http_requests_total[5m])

Pod Resource Usage:

container_memory_usage_bytes{pod=~"my-app-.*"}

Time Ranges

Supported time range values:

  • 1h - Last hour
  • 24h - Last 24 hours
  • 7d - Last 7 days
  • 30d - Last 30 days

Query Steps

For range queries, you can specify the resolution step:

  • 30s - 30 second intervals
  • 1m or 60s - 1 minute intervals (default)
  • 5m - 5 minute intervals
  • 1h - 1 hour intervals

Events Module

Query Kubernetes events with improved parameter names:

  • get_pod_events - Use pod parameter for specific pod name
  • get_deployment_events - Use deployment parameter for specific deployment name
  • get_nodes_events - Use node parameter for specific node name

Quick Start

  1. Configure your monitoring endpoints in configs/config.yaml
  2. Set sensitive tokens via environment variables
  3. Run the server in stdio mode:
./bin/ops-mcp-server --config configs/config.yaml

Building

go build -o bin/ops-mcp-server cmd/server/main.go