Traces
The trunk line. End-to-end waterfalls across every service, one ID.
Your coding agent sets up Traceway and debugs production for you.
MIT-licensed. Self-host in minutes or start free in the cloud.
Open-source APM. Every hop is a station. Logs, metrics, and replay are transfers. You keep your place.
On the same trace
The trunk line. End-to-end waterfalls across every service, one ID.
Where the line breaks. Grouped, ranked, sitting on the span that died.
Transfer to the browser. The click that sent them down this path.
Structured, trace-linked, sub-second search. Same ticket.
Host, runtime, and custom charts next to the station they belong to.
LLM cost, tokens, conversations. Same OTLP pipe as everything else.
Why it matters
Traceway catches the error, the session replay, and the exact failing span before they close the tab.
Detect
Every endpoint ranked by five service-level signals. Alerts to Slack, GitHub, or a webhook. The 412-event cart null jumps the queue.
Resolve
Open the exception and the session is already there, source-mapped, tied to the span that died. Context-switching is the bug.
Agent skills
One command installs the skills into Claude Code, Cursor, Codex, or any agent that reads SKILL.md. From there it queries exceptions, logs, and metrics itself.
npx skills add tracewayapp/tracewayCommunity
MIT licensed and developed in public. Star the repo, file issues, and help shape the roadmap. No BSL. No open-core bait.
Your data. Your metal.
Full Cloud parity. Point OTLP at it and you’re in. No collector required, no vendor SDK.
$ git clone github.com/tracewayapp/traceway $ cd traceway && docker compose up -d # clickhouse · postgres · collector ✓ dashboard at http://localhost:3000
Pricing that doesn’t lie
Two meters, agreed up front: exceptions and ingested data. Requests, seats, and hosts never touch the bill — and the whole line runs self-hosted for free.
Start for free. Self-host whenever you want. Book a demo if you’d like a walkthrough.