Three pillars. Each one solves a problem most observability vendors gave up on.
Most platforms downsample after 5 minutes. We don’t. Stream millions of events per second from your fleet and query the last 60 seconds with sub-100ms latency. Your dashboards stop lying to you about what’s actually happening right now.
Built on a custom columnar engine. Self-hostable for sovereignty-sensitive workloads. Single binary, zero dependencies.
Read the engineering deep-dive →Forget 200-line YAML alert configs. Orbital learns your fleet’s behavioral baseline in 48 hours, then surfaces deviations as plain-English summaries. “Checkout latency p99 grew 4× on eu-west since deploy 42a3f1 — likely caused by the new stripe.session.create retry path.”
Routes to Slack, PagerDuty, MS Teams, or webhooks. Auto-attaches the relevant trace + git diff. Three clicks to acknowledge or escalate.
See an example anomaly →# Anomaly detected (confidence 96%) service: checkout-api region: eu-west-1 metric: latency_p99 deviation: +412% over 2 min started: 3 min after deploy // likely cause stripe.session.create retry path now blocking on idempotency-key lookup. // suggested action git revert 42a3f1
OQL feels like SQL but reads like English. Auto-complete every field. Diff two windows in one keystroke (DIFF NOW vs 1h AGO). Share any query as a URL — your team’s observability becomes conversation, not tribal knowledge.
Cheat-sheet, full grammar, and an interactive playground are linked in the docs. Most engineers pick it up in 20 minutes.
Open the OQL playground →DIFF p99(latency_ms) FROM api BETWEEN now() AND deploy() vs p99(latency_ms) FROM api BETWEEN deploy() AND deploy(-1) GROUP BY endpoint SORT DESC;
30-day free trial. No credit card. 4-minute setup.