Your recommendation service just went live in production. You need alerts configured now: high latency, error spikes, and service availability. But manually building alert rules means clicking through query builders, setting thresholds, and configuring notification channels.
You open your AI assistant, connected to SigNoz via the MCP server, and describe what you need in plain English.
Prerequisites
- Connect your AI assistant to SigNoz using the MCP Server guide.
- Make sure your services are instrumented with distributed tracing. See Instrument Your Application if you haven't set this up.
Step 1: Create Alerts from a Natural Language Prompt
Confirm the recommendation service is sending data. Then create three alerts: one for p99 latency above 2 seconds, one for error rate above 5%, and one for service availability.
Your assistant will first verify the service exists and is actively sending telemetry data, then create the alerts:
Using "demo-alerts" notification channel for all three alerts. Creating them now.
Alert 1: p99 latency above 2 seconds
✓ Created
Alert 2: error rate above 5%
✓ Created
Alert 3: Service availability — using alertOnAbsent to fire when no spans are received from the service.
✓ Created
All done. Here's the summary:
Recommendation service status: Sending data — 87 calls in the last hour (~0.024/sec), currently running at ~5.7% error rate.
Alerts created, all routed to demo-alerts, 5m evaluation window / 1m frequency:
1. P99 Latency > 2s (warning) — fires when recommendation's p99 trace duration exceeds 2000ms, recovers below 1500ms.
2. Error Rate > 5% (critical) — computed as errors / total_calls * 100 on recommendation traces, recovers below 3%.
3. No Data / Availability (critical) — fires if no spans arrive from the service for 5 minutes.
Two things worth flagging:
• Heads up on the error rate alert: the service is currently at ~5.7%, so this alert may fire immediately once evaluation starts. Let me know if you'd like the threshold raised or the window lengthened.
• I used a 5-minute absentFor for the availability alert. If you want something more sensitive (e.g., 2 minutes) or less noisy (e.g., 10 minutes), happy to adjust.
The alerts are now active and will notify you when thresholds are breached.
Final Summary
You now have three production-ready alerts configured from a simple prompt. Your recommendation service is being monitored for latency issues, error spikes, and availability problems. All without manually configuring query builders or notification channels.
Under the Hood
During this workflow, the MCP server called these tools:
| Step | MCP Tool | What It Did |
|---|---|---|
| 1 | signoz_list_services | Verified the recommendation service exists and is actively sending telemetry data |
| 1 | signoz_create_alert | Created the p99 latency alert with threshold and notification configuration |
| 1 | signoz_create_alert | Created the error rate alert with percentage-based threshold |
| 1 | signoz_create_alert | Created the service availability alert monitoring for data absence |
Related Use Cases
- Post Deployment Monitoring - Compare key metrics before and after a deployment to detect performance regressions or unexpected changes.
- Alert Correlation Analysis - When multiple services alert simultaneously, identify whether it's a cascade from one failure or separate incidents.
- Dashboard Creation from Natural Language - Create custom dashboards by describing what you want to visualize in plain English.
If you need help with the steps in this topic, please reach out to us on SigNoz Community Slack.
If you are a SigNoz Cloud user, please use in product chat support located at the bottom right corner of your SigNoz instance or contact us at cloud-support@signoz.io.
