Traceway

WHAT IS IT?

An OpenTelemetry-native observability platform that groups into one product what you usually assemble from Prometheus, Loki, Tempo and a paging tool: logs, traces, metrics, exceptions, session replay, LLM tracking and on-call. The backend is Go (Gin), the frontend is SvelteKit, ingestion goes through OTLP/HTTP, and the license is MIT, with no BSL and no open-core split.

WHY IS IT INTERESTING?

• Everything is correlated: span waterfalls with direct navigation to the trace's logs, per-endpoint metrics (P50/P95/P99, ranked by Apdex) and flame graph profiling (CPU, heap, goroutines) with version-to-version diffing.
• Deployment options: docker compose with ClickHouse and PostgreSQL for a real production setup, or a single container in embedded SQLite/DuckDB mode with zero external dependencies. A Go application can even import it as a library.
• Readable exceptions: stack trace grouping by fingerprint, source map resolution (webpack, esbuild, Vite) and mobile crash symbolication, with an engine written in pure Go.
• Frontend and mobile included: browser and Flutter session replay, React, Vue and Svelte SDKs on the web side and Android, iOS, React Native on mobile, on top of classic server-side OTel.
• Built-in on-call: rotation schedules, escalation policies and notifications over email, Slack, SMS, Telegram or Pushover, with no separate product to wire in.

USE CASES

• Replace a Prometheus + Loki + Tempo + paging stack with a single service to operate.
• Observability for a side project or a small production: the embedded SQLite/DuckDB mode fits in one container.
• Track cost, tokens and latency of an AI app's LLM calls in the same place as its traces.
• Let an agent do the work: Claude Code/Cursor skills ship with it, /traceway-setup to instrument the project and /traceway to query production data.

If self-hosting is not an option, a managed cloud version runs at cloud.tracewayapp.com.