Prometheus MCP Server (Go)

A Model Context Protocol (MCP) server for Prometheus integration. Native Go binary with built-in Kubernetes connectivity via client-go.

Installation

npx

npx -y mcp-prometheus@latest

MCP Client Configuration

Add to your MCP client configuration (VS Code, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "prometheus": {
      "command": "npx",
      "args": ["-y", "mcp-prometheus@latest"],
      "env": {
        "PROMETHEUS_URL": "http://localhost:9090"
      }
    }
  }
}

VS Code

code --add-mcp '{"name":"prometheus","command":"npx","args":["-y","mcp-prometheus@latest"],"env":{"PROMETHEUS_URL":"http://localhost:9090"}}'

Kubernetes Auto-Connect

Automatically connects to Prometheus running in OpenShift/Kubernetes via the K8S API service proxy. Uses native kubeconfig/in-cluster config via client-go. No kubectl or port-forwarding required.

Default: openshift-monitoring/prometheus-operated:9090

{
  "mcpServers": {
    "prometheus": {
      "command": "npx",
      "args": ["-y", "mcp-prometheus@latest"]
    }
  }
}

Binary

Download from GitHub Releases or build from source:

make build
./mcp-prometheus

---

Configuration

Environment Variables

Variable	Description
PROMETHEUS_URL	Direct Prometheus API URL (overrides K8S auto-connect)

CLI Flags

Flag	Description	Default
--url	Direct Prometheus URL	-
--namespace	Kubernetes namespace	openshift-monitoring
--service	Kubernetes service name	prometheus-operated
--service-port	Kubernetes service port	9090
--service-scheme	Service scheme (http/https)	https
--kubeconfig	Path to kubeconfig file	auto-detect

Precedence: --url / PROMETHEUS_URL > K8S auto-connect

Connection strategy:

1. Direct URL (if --url or PROMETHEUS_URL is set)
2. K8S API proxy (auto-detect kubeconfig or in-cluster config)

---

Tools (11)

Basic Queries

Tool	Description
query	Execute a PromQL instant query
queryRange	Execute a PromQL range query over time
getTargets	Get scrape targets status
getRules	Get alerting and recording rules
getPrometheusStatus	Get server version and runtime info

Cluster Diagnostics

Tool	Description
getClusterHealthOverview	Comprehensive cluster health overview
diagnoseNode	Diagnose a specific node's health
diagnoseNamespace	Diagnose a namespace's health

Resource Analysis

Tool	Description
getTopResourceConsumers	Top CPU/memory/network consumers
investigatePod	Deep investigation of a specific pod
compareTimeRanges	Compare metrics between two time periods

---

Example Prompts

"What's the current CPU usage across all nodes?"
"Show me the cluster health overview"
"Which pods are using the most memory?"
"Diagnose node worker-1"
"Are all scrape targets healthy?"
"Compare CPU usage now vs 24 hours ago"
"Investigate pod my-app in namespace production"
"What alerting rules are defined?"

---

Development

Build

make build              # Build for current platform
make build-all-platforms # Cross-compile for all platforms

Container

`

... View full README on GitHub