import { z } from 'zod';
import { log } from '../core/logger.js';
/**
* MCP Prompts for Datadog workflows.
*
* Prompts provide pre-built conversation templates that guide
* LLMs through common Datadog operations and investigations.
*/
export function registerPrompts(server) {
// --- Investigate Monitor ---
server.prompt(
'investigate-monitor',
'Investigate a Datadog monitor that is alerting or has issues. Provides a structured investigation workflow.',
{
monitor_id: z.string().describe('The monitor ID to investigate'),
},
async ({ monitor_id }) => ({
messages: [
{
role: 'user',
content: {
type: 'text',
text: [
`Investigate Datadog monitor ${monitor_id}. Follow this workflow:`,
'',
'1. First, use the get_monitor tool with this monitor_id to retrieve the monitor details.',
'2. Check the monitor state (OK, Alert, Warn, No Data).',
'3. Review the monitor query and thresholds.',
'4. If the monitor is alerting:',
' - Use search_logs to look for related errors in the last hour.',
' - Use query_metrics to check the underlying metric trend.',
' - Check if there are any related downtimes with list_downtimes.',
'5. Summarize findings with:',
' - Current state and when it changed',
' - Likely root cause',
' - Recommended actions',
].join('\n'),
},
},
],
})
);
// --- Create Dashboard ---
server.prompt(
'create-dashboard',
'Guided workflow for creating a new Datadog dashboard with best practices.',
{
purpose: z.string().describe('What this dashboard is for (e.g., "API performance monitoring")'),
service: z.string().optional().describe('Service name to focus on'),
},
async ({ purpose, service }) => ({
messages: [
{
role: 'user',
content: {
type: 'text',
text: [
`Help me create a Datadog dashboard for: ${purpose}`,
service ? `Focused on service: ${service}` : '',
'',
'Follow this workflow:',
'1. First, list existing dashboards with list_dashboards to avoid duplicates.',
'2. Suggest a dashboard structure with appropriate widgets:',
' - Summary/overview widgets at the top',
' - Key metrics timeseries in the middle',
' - Detailed breakdown widgets at the bottom',
'3. For each widget, suggest appropriate:',
' - Widget type (timeseries, query_value, toplist, heatmap, etc.)',
' - Metric queries with proper aggregation',
' - Meaningful titles and display options',
'4. Create the dashboard using create_dashboard with layout_type "ordered".',
'5. Share the dashboard URL.',
].join('\n'),
},
},
],
})
);
// --- Analyze Logs ---
server.prompt(
'analyze-logs',
'Guided log analysis workflow for investigating issues in Datadog logs.',
{
query: z.string().describe('Initial search query or error description'),
timeframe: z.string().optional().describe('Time range (e.g., "1h", "24h", "7d"). Default: 1h'),
},
async ({ query, timeframe }) => {
const tf = timeframe || '1h';
return {
messages: [
{
role: 'user',
content: {
type: 'text',
text: [
`Analyze Datadog logs for: ${query}`,
'',
`Time range: last ${tf}`,
'',
'Follow this workflow:',
`1. Use search_logs with query "${query}" and from "now-${tf}" to find matching logs.`,
'2. Analyze the results:',
' - Count error frequency and patterns',
' - Identify common error messages',
' - Note affected services/hosts',
'3. If errors are found, drill deeper:',
' - Refine the search query with specific error patterns',
' - Check for correlated events around the same timeframe',
' - Use query_metrics to check if related metrics spiked',
'4. Provide a summary with:',
' - Total log count and error rate',
' - Top error patterns',
' - Affected components',
' - Recommended next steps',
].join('\n'),
},
},
],
};
}
);
// --- Incident Response ---
server.prompt(
'incident-response',
'Structured incident response workflow using Datadog data for diagnosis and triage.',
{
description: z.string().describe('Brief description of the incident'),
severity: z.enum(['SEV-1', 'SEV-2', 'SEV-3', 'SEV-4']).optional().describe('Incident severity level'),
},
async ({ description, severity }) => ({
messages: [
{
role: 'user',
content: {
type: 'text',
text: [
`Incident Response: ${description}`,
severity ? `Severity: ${severity}` : '',
'',
'Follow this incident response workflow:',
'',
'## 1. Assess Current State',
'- Use list_monitors with group_states "alert,warn" to find all alerting monitors.',
'- Use list_incidents to check for existing incidents.',
'- Use list_hosts to verify infrastructure health.',
'',
'## 2. Gather Evidence',
`- Use search_logs with query related to "${description}" in the last hour.`,
'- Use query_metrics to check key infrastructure metrics (CPU, memory, network, errors).',
'- Check for any recent downtimes with list_downtimes.',
'',
'## 3. Determine Impact',
'- List affected services, hosts, and users.',
'- Quantify error rates and latency increases.',
'- Identify the blast radius.',
'',
'## 4. Recommend Actions',
'- Immediate mitigation steps.',
'- Communication plan.',
'- Follow-up investigation items.',
'',
'## 5. Document',
'- Timeline of events.',
'- Root cause hypothesis.',
'- Actions taken.',
].join('\n'),
},
},
],
})
);
// --- Service Health Check ---
server.prompt(
'service-health-check',
'Comprehensive health check for a specific service using Datadog monitoring data.',
{
service_name: z.string().describe('Name of the service to check'),
},
async ({ service_name }) => ({
messages: [
{
role: 'user',
content: {
type: 'text',
text: [
`Perform a comprehensive health check for service: ${service_name}`,
'',
'Check the following areas:',
'',
'## 1. Monitors',
`- Use list_monitors with name "${service_name}" to find all related monitors.`,
'- Report any monitors in alert or warn state.',
'',
'## 2. Logs',
`- Use search_logs with query "service:${service_name} status:error" from the last hour.`,
'- Report error count and top error messages.',
'',
'## 3. Metrics',
'- Use query_metrics to check key service metrics:',
` - Request rate: avg:trace.http.request.hits{service:${service_name}}`,
` - Error rate: avg:trace.http.request.errors{service:${service_name}}`,
` - Latency: avg:trace.http.request.duration{service:${service_name}}`,
'',
'## 4. Infrastructure',
`- Use list_hosts with filter "${service_name}" to check host health.`,
'',
'## 5. Summary',
'- Overall health: Healthy / Degraded / Critical',
'- Key issues found',
'- Recommended actions',
].join('\n'),
},
},
],
})
);
log('Registered 5 MCP prompts');
}
