Distributed tracing
Follow a single request across dozens of services with waterfall views, flame graphs, and automatic span correlation.
Tracelight streams distributed traces, metrics, and logs into one query surface — so your engineers ship faster and sleep through the night.
$ tracelight tail --service checkout-api --live 200 POST /v1/orders 42ms trace=a91f…c2 200 GET /v1/cart/items 18ms trace=7b02…9d 429 POST /v1/payments 311ms trace=c5d1…10 # rate-limited 200 GET /v1/inventory/sync 27ms trace=e4aa…3f 503 POST /v1/shipping/quote 1.2s trace=08fc…b7 # upstream timeout › anomaly detected — p99 latency +214% on shipping/quote
Follow a single request across dozens of services with waterfall views, flame graphs, and automatic span correlation.
High-cardinality metrics with sub-second aggregation. No pre-defined dashboards required — just ask.
Baseline-aware detection surfaces p99 spikes, error-rate drifts, and saturation before your users notice.
Jump from a slow span straight to the exact log lines that explain it — no tab hopping, no grep spelunking.
Drop-in OTLP ingestion. Point your existing collectors at Tracelight and see data flowing in minutes.
Per-service ingest budgets and sampling controls keep your observability bill predictable at any scale.
Stop building brittle dashboards. Write a query, save it as a monitor, share it as a link. Your whole team speaks the same observability language.
See it on your data{ service = "checkout-api" && duration > 300ms } | by(route) | p99(duration) → /v1/shipping/quote 1.24s → /v1/payments 0.31s → /v1/orders 0.09s
“We cut mean-time-to-resolution from 47 minutes to under 6. Tracelight is the first tab our on-call engineers open.”
“Setup took fifteen minutes and we had real traces flowing before our next standup. No SDK gymnastics required.”
“TraceQL replaced eleven brittle dashboards. Now every on-call engineer just writes the query they actually need.”
“We cut mean-time-to-resolution from 47 minutes to under 6. Tracelight is the first tab our on-call engineers open and the last one they close.”
Tell us where your stack lives and we'll spin up a workspace with sample data so you can explore before you connect a single collector.