Observability

See everything.
Get paged when it breaks.

Events, metrics and alarms for everything you run on koigrid — a timeline of what happened, health metrics for your databases and caches, and threshold alarms by email. The built-in, flat-rate alternative to AWS CloudWatch.

CloudWatch is another product to wire up — and pay for

On AWS, logs, metrics, dashboards and alarms are a separate service billed per metric, per alarm, per GB of logs and per API call — and you still connect each resource to it. koigrid ships observability with every resource: a per-org event feed, real health metrics sampled from your databases and caches, and alarms — no extra service to configure, at a flat price.

How koigrid compares

koigridAWS CloudWatch
Event timeline per tenantBuilt-inLogs Insights (extra)
Resource health metricsSampled automatically (DB/Redis)Per-metric, you wire it
Threshold alarms → emailBuilt-inAlarms + SNS (extra)
Consumption metrics + costBuilt-inCost Explorer (separate)
SetupAutomatic with every resourcePer-resource, per-metric
PricingIncluded / flatPer metric + alarm + GB logs + API
Query by API or AI agentNativeAPI (complex)

From public docs, 2026. koigrid samples DB/Redis health (connections, replication lag, memory, disk) into a time series and evaluates alarms on a tick.

What you actually watch in production

Event timeline

Every action, request and error for your organization — filter by type, severity, entity or time. Errors carry a citable errorRef.

Health metrics

Connections and replication lag for databases, memory and clients for caches, disk usage — sampled automatically into a time series.

Alarms by email

Alarm on consumption or on a resource-health metric (Redis memory > 80%, replication lag, disk full) — notified by email when it trips.

Consumption + cost

Daily series of CPU, memory, egress and requests, with a synthetic cost metric — no separate Cost Explorer.

Agent-native

Query events, metrics and alarms from Claude Code or any LLM via the API — your agent can watch its own workloads.

Nothing breaks in silence

The platform itself is monitored end-to-end; failures are caught and surfaced, not swallowed.