Smart Alarms
Know the
why, not just what.
Agents correlate context across every chart before alerting. Not static thresholds — contextualized intelligence that explains why an alarm fired, traces root causes, and proposes what to do next.
Describe conditions in plain English. Deploy across one or more charts. Get alerts with full context.
Alarm Types
Three alarm types, infinite combinations.
Every alarm in GlacierHub runs through an agent that understands your data. Set boundaries on a single metric, correlate conditions across multiple charts, or describe complex rules in natural language — the system monitors and alerts with full context.
Single-chart thresholds
Set min/max bounds on any metric. The chart’s agent monitors the data stream and fires when values breach the range — with context on what likely caused it. Not just “revenue dropped 12%” but “revenue dropped 12% as Stripe reported declining conversion rates.”
Cross-chart correlation
Compound conditions across multiple chart agents working together via the orchestrator. “Alert when API latency spikes AND deploy frequency is above normal.” The orchestrator coordinates agents to evaluate conditions no single source could detect. Root cause analysis across your entire dashboard.
Natural language rules
No rule builders or config files. Describe alarm conditions in plain English and agents translate them into monitored triggers. “Notify me when daily active users drop by more than 10% compared to last week’s average.” The system understands, monitors, and alerts with contextual explanations when conditions match.
Beyond static thresholds.
Context-aware: Agents explain WHY an alarm fired, not just that it fired. Revenue dropped because conversion rates declined, not just “revenue is down.” The chart agent traces the event back through its data sources and related metrics to identify the root cause.
Root cause analysis: When an alarm triggers, the agent doesn’t stop at detection. It investigates: What changed in the upstream data? Are related metrics also affected? Did a recent deployment correlate with the anomaly? The analysis is delivered with the alert.
False positive reduction: Cross-referencing related metrics before alerting. A spike in error rates might look alarming until the agent checks deploy logs and confirms you just shipped a new feature with expected log verbosity. Context prevents noise.
Suggested actions: Agents propose what to do when an alarm fires. “Revenue dropped 12% while ad spend increased 8%. Consider pausing the underperforming campaign and reallocating budget to the top-performing segment.” Actionable intelligence, not just alerts.
Revenue dropped 12% while ad spend increased 8% over the last 7 days. The Stripe agent detected declining conversion rates from 3.2% to 2.1%; the Google Ads agent confirmed rising CPC from $2.40 to $3.80 in the same period.
Consider pausing the “Brand Awareness Q1” campaign (CPC +$1.40, CVR −35%) and reallocating budget to “Product Launch” (CPC +$0.20, CVR −8%). Projected recovery: +$4,200 daily revenue.
Notification Channels
Alerts wherever you work.
Configure notification channels once, apply them to any alarm. Every alert includes full context: why it triggered, what data changed, and suggested next steps.
Slack
Real-time channel notifications with full alarm context. Thread replies to acknowledge alerts and track resolution directly in Slack.
Digest or instant alerts with full analysis. HTML-formatted reports include charts, suggested actions, and links to investigate further.
Webhooks
POST to any endpoint for custom integrations. Trigger PagerDuty, OpsGenie, or your own incident management system with structured payloads.
In-app
Notification center in the GlacierHub dashboard. Filter by severity, acknowledge in bulk, and jump directly to the chart that triggered.
Alarm Builder
Natural language alarm creation.
Natural language input: Type or speak your alarm condition. “Alert me when daily signups drop below 100 and email open rate is under 15%.” The system parses metrics, operators, and thresholds automatically.
Visual rule builder: For power users who prefer explicit control. Click to add conditions, chain with AND/OR logic, set time windows and cooldowns. Every option exposed without writing code.
Condition chaining: Combine single-chart and cross-chart conditions with boolean logic. “(Revenue < $10k OR Churn > 5%) AND Deploy Count > 3.” Unlimited nesting, arbitrary complexity.
Time windows and cooldowns: Prevent flapping with configurable windows. “Only trigger if the condition persists for 15 minutes.” Set cooldown periods to avoid repeated alerts for the same issue.
Severity levels: Low, Medium, High, Critical. Route different severities to different channels. Critical goes to PagerDuty, Medium to Slack, Low to email digest.
Alarm Builder
Describe condition
Alert when daily signups drop below 100 and email open rate is under 15%
Conditions
Severity
Notification Channels
Alarm History
4 minutes ago
Revenue dropped 12% while ad spend increased 8%
2 hours ago by @sarah
p95 latency exceeded 200ms threshold
Yesterday at 3:42 PM
Active connections dropped to 85% of pool limit
2 days ago
Churn rate exceeded 5% while NPS fell below 40
Full audit trail.
Complete history: Every alarm that fired, when it triggered, who acknowledged, and what action was taken. Searchable, filterable, exportable. Never lose context on what went wrong and how it was resolved.
Context snapshots: The system captures what the data looked like when the alarm triggered — not just the metric value but all related charts, recent deploys, and correlated events. Investigate retroactively with the full picture.
Resolution tracking: Acknowledge alarms, assign to team members, track time-to-resolution. Add notes about what fixed the issue. Build institutional knowledge about recurring patterns and effective responses.
Pattern detection: The system identifies recurring alarms and flapping detection (alarms that trigger and resolve repeatedly). Suggest raising thresholds or adding cooldown periods to reduce noise without missing real issues.
Performance
Built for scale and speed.
Sub-second detection, unlimited alarm rules, and real-time monitoring across your entire dashboard. Set conditions in natural language and deploy instantly.
Contextual
Agents explain why alarms fired, not just that they did. Root cause analysis included with every alert.
Cross-chart
Compound conditions across multiple chart agents working together via the orchestrator.
Natural language
Describe alarm conditions in plain English. Agents parse and monitor automatically.
Sub-second
Detection latency from data ingestion to alert dispatch. Real-time monitoring at scale.
Unlimited
Alarm rules per dashboard. Set as many conditions as you need without hitting quotas.
4 channels
Slack, Email, Webhooks, In-app. Route different severity levels to different destinations.
Intelligent Features
Intelligence built into every alarm.
Agents don’t just monitor thresholds. They analyze patterns, reduce false positives, correlate events, and propose actions. Every alarm is an opportunity for insight, not just noise.
False positive reduction
Agents cross-reference related metrics before alerting. A spike in error rates correlated with a recent deploy might be expected. Context prevents alert fatigue.
Root cause tracing
The agent traces events back through data sources. Revenue dropped because conversion rates fell, which correlates with rising ad costs and declining landing page performance.
Pattern detection
Identifies recurring alarms and suggests adjustments. If an alarm triggers every Monday morning, the system proposes raising thresholds or excluding that time window.
Suggested actions
Agents propose what to do when an alarm fires. Pause an underperforming campaign, scale infrastructure, investigate a specific cohort. Actionable intelligence, not just alerts.
Flapping detection
Identifies alarms that trigger and resolve repeatedly. Suggests adding cooldown periods or adjusting thresholds to reduce noise without missing real issues.
Team coordination
Assign alarms to team members, track acknowledgment, and measure time-to-resolution. Build runbooks around recurring patterns. Turn alarms into institutional knowledge.
Get Started
Never miss what matters
Set up context-aware alarms in minutes. Let agents monitor, correlate, and alert with intelligence.
Get Early Access