Skip to main content
Platform Reference

Send Traces to Your Own Collector

Point indie.money at an OTLP collector and every run arrives as a trace, with one child span per node.

Written By Philippe

Last updated 3 days ago

Every run your workflows do can be exported to your own observability stack. Youpoint the platform at an OTLP endpoint you control, and each execution arrives asa trace you can search, alert on, and keep for as long as your own retention says.

What gets sent

One span per run, carrying the outcome (success or failure), how long it took,what it cost, which trigger started it, and — when a run fails — which nodefailed. Inside that span sits one child span per node, with its own wall-clocktiming, so a slow run shows you where the time went rather than just how long ittook in total.

Setting it up

Open your producer dashboard and find the trace export card. You need threethings:

  • Endpoint — the base URL of your collector, for examplehttps://otel.example.com. We add the signal path. Pasting the full traces URLworks too. It must be https; plain http would put your collector token on thewire in the clear, so it is rejected.
  • Auth header name — usually Authorization, but your vendor may want itsown header.
  • Token — the credential your collector expects. It is stored the same wayyour workflow credentials are, and is never shown back to you afterwards.

Saving verifies. When you save, we send a single span to your endpoint and show youthe result immediately, so you find out then and there whether the address and tokenare right — instead of discovering it the next time something breaks. Use Send testspan afterwards to re-check a sink you have already saved.

Things worth knowing

  • Private, loopback, and internal addresses are refused. The collector has to bereachable from the public internet.
  • Trace export is additive. Turning it on changes nothing about how your runsexecute, what they cost, or what you see in your own dashboard.
  • Removing the configuration stops the export immediately. Traces alreadydelivered to your collector are yours and stay where they are.
  • Average response time is already answerable from the trace itself: every run's spancarries its start and end, so your collector can chart mean and percentile durationwith no extra setup.
  • One sink covers every workflow you own. Removing it stops trace export for all of them.

Was this helpful?

Still need help? Ask the team