mcp-prometheus mcp

v1.0.1 · MCP Tool · monitoring · registry.pascalai.org

Prometheus HTTP API v1 — PromQL queries, targets, alerts, rules and metadata.

prometheusmetricsmonitoringobservabilitydevopspromql

Input Parameters

Parameter	Type	Description
operationrequired	string	Operation to perform. One of: `query`, `query_range`, `labels`, `label_values`, `series`, `targets`, `alerts`, `rules`, `metadata`, `tsdb_status`.
baseUrlrequired	string	Prometheus base URL (default: http://localhost:9090). Default: `http://localhost:9090`.
usernameoptional	string	Basic auth username.
passwordoptional	string	Basic auth password.
bearerTokenoptional	string	Bearer token for auth (takes priority over username/password).
queryoptional	string	PromQL query expression e.g. up, rate(http_requests_total[5m]).
startoptional	string	Start timestamp (RFC3339 or Unix timestamp) for query_range.
end_optional	string	End timestamp (RFC3339 or Unix timestamp) for query_range.
stepoptional	string	Query resolution step for query_range e.g. 15s, 1m, 5m. Default: `15s`.
timeoptional	string	Evaluation timestamp for instant query (RFC3339 or Unix, default: now).
labelNameoptional	string	Label name for label_values operation.
matchoptional	string	Series selector for series/label_values e.g. {job="prometheus"}.
timeoutoptional	string	Evaluation timeout e.g. 30s (passed to Prometheus).

Output Fields

Field	Type	Description
status	string
data	object
error	string
errorType	string

Examples

Query CPU usage

// Input
{
  "operation": "query",
  "baseUrl": "http://prometheus:9090",
  "query": "100 - (avg(rate(node_cpu_seconds_total{mode='idle'}[5m])) * 100)"
}

// Output
{
  "status": "success",
  "data": {}
}

List active alerts

// Input
{
  "operation": "alerts",
  "baseUrl": "http://prometheus:9090"
}

// Output
{
  "status": "success",
  "data": {}
}

Install & Discovery

Install

ppm install mcp-prometheus

Get JSON Schema

GET /v1/packages/mcp-prometheus/1.0.1/schema

Discover by keyword

GET /v1/mcp/discover?q=prometheus

Discovery hint: Install with ppm install mcp-prometheus or invoke remotely via POST /v1/invoke/mcp-prometheus on the MCP Service.

PascalAI Usage

uses toolslib;
var Tool := LoadTool('mcp-prometheus');
var R := Tool.Call(JsonObj(['operation','query','baseUrl','http://localhost:9090','query','up']));
Writeln(R.ToJSON);