Beacon stitches every trace, metric, and log into a single timeline and walks you from a red SLO to the exact span, query, and line of code behind it — median root-cause in 4 minutes, not 4 hours.
$ beacon trace 9f31c2 --explain
trace 9f31c2 POST /api/checkout status=500 p99=4.91s
├─ api-gateway 12ms ok
├─ checkout-svc 18ms ok
└─ payments-svc 4.83s ERROR ← 98% of latency
└─ db.query SELECT * FROM ledger no index on tenant_id
[anomaly] payments-svc p99 +1,840% vs 7d baseline
[deploy] correlated to release v2.31.0 (14m ago, @priya)
[culprit] payments/ledger.go:142 full-table scan added in #4471
→ rollback v2.31.0 or open PR with composite index?
✓ alert auto-resolved when p99 < 250msTrusted by on-call engineers at teams that ship every day
No more bouncing between a tracing tool, a metrics tool, and a log search at 3 a.m. Beacon correlates all three on a shared timeline and connects every spike to the deploy that caused it.
Tail-based sampling keeps the slow and failed traces, drops the boring ones, and waterfalls every request across all your services. Click any span to jump straight to its logs, the exact SQL it ran, and the commit that shipped it. Trace search returns p99 in under 200ms over a billion spans.
Slice by user_id, region, or feature flag without pre-aggregation. PromQL-compatible, 13-month retention, sub-second queries.
Every log line carries its trace_id, so one click pivots from a noisy stream to the precise request it belongs to.
Define an SLO in a line of YAML. Beacon burns the budget, alerts before you breach, and shows exactly what's eating it.
Anomaly detection ties every regression to the release, flag, or config change that introduced it — no static thresholds to babysit.
What on-call actually feels
Point your existing OTel exporters at Beacon and you're done. No proprietary agent, no rip-and-replace, no vendor lock-in on the data you already collect.
Send OpenTelemetry traces, metrics, and logs over the wire you already use. Existing SDKs and collectors work unchanged.
A single `helm install` deploys collection across your cluster. First traces land in under two minutes.
Run Beacon in your own VPC or our single-tenant cloud. Export raw OTLP out at any time — no walled garden.
The full query engine ships as a CLI. Page through traces, tail logs, and check SLOs without leaving your shell.
Per-gigabyte pricing punishes you for instrumenting more. We charge for the hosts you run, so your bill never spikes with your traffic.
For side projects and small teams getting started.
For engineering teams that live on-call.
For regulated, high-volume, multi-region fleets.
No. Beacon speaks native OpenTelemetry over OTLP. Point your existing exporters or collector at our endpoint and your current traces, metrics, and logs start flowing — no proprietary agent, no code rewrite.
We never charge per gigabyte. Pricing is a flat rate per host, so you can crank up sampling and cardinality without watching the meter. Tail-based sampling also keeps the traces that matter and drops the noise before it ever costs you.
About two minutes. A single `helm install` deploys collection across your cluster, and the first traces and SLO dashboards populate before the rollout finishes. Most teams resolve their first real incident on Beacon the same day.
Yes. Run Beacon fully inside your VPC or in our isolated single-tenant cloud, pin data residency by region, and export raw OTLP back out whenever you like. Your telemetry never co-mingles with another customer's.
Beacon pushes deploy-aware, budget-burn alerts straight into PagerDuty, Opsgenie, and Slack, and exposes a PromQL-compatible query API so your existing Grafana panels and runbooks keep working.
Team and Enterprise plans keep 13 months of full-resolution traces and metrics — no downsampling, no cold-storage tax. You can query last year's incident as fast as last night's.