Skip to main content
n8n can export workflow and node execution traces over OTLP. Recent n8n versions can also emit agent spans using OpenTelemetry GenAI semantic conventions. Send these spans to Phoenix to inspect workflow latency, node failures, and agent behavior. No workflow changes are required — you configure n8n with environment variables.
n8n OpenTelemetry tracing is marked preview in n8n’s documentation. OTEL tracing is available from n8n 2.19.0, UI-based OTEL configuration is available from 2.27.0, and agent tracing is available from 2.33.0. Validate the behavior on your n8n version before relying on it for production operations.

​
Prerequisites

  • A self-hosted n8n instance with OpenTelemetry tracing support. Use n8n 2.19.0 or later for workflow tracing and 2.33.0 or later for agent tracing.
  • A running Phoenix instance that your n8n processes can reach.

​
Launch Phoenix

The fastest way to get started with Phoenix is by signing up for a free Phoenix Cloud account. If you prefer, you can also self-host Phoenix next to n8n. Note your Phoenix endpoint — for example http://localhost:6006 for a local instance or https://app.phoenix.arize.com/s/<your-space-name> for Phoenix Cloud — and, if authentication is enabled, an API key from the settings page.

​
Configure n8n

Set these variables on each n8n process that should emit traces, including main, worker, and webhook processors in queue mode:
If your Phoenix instance has authentication enabled (for example, Phoenix Cloud), set the endpoint to your instance URL and pass your API key as a header:
Restart n8n after setting the environment variables.
n8n appends N8N_OTEL_EXPORTER_OTLP_TRACING_PATH to N8N_OTEL_EXPORTER_OTLP_ENDPOINT. Set the endpoint to the base Phoenix URL and the tracing path to /v1/traces. Keep the default http/protobuf protocol: Phoenix receives OTLP over HTTP on the same port as its UI. When n8n runs in a container and Phoenix runs on the host, use a hostname the container can resolve, such as http://host.docker.internal:6006.
OTEL_RESOURCE_ATTRIBUTES sets the Phoenix project through the openinference.project.name resource attribute. Without it, n8n traces land in the default project. For production deployments, you can use n8n’s _FILE variants for sensitive values when your deployment platform mounts secrets as files.

​
Run n8n in a container

This example runs n8n with tracing pointed at a Phoenix instance on the host:

​
Enable n8n agent tracing

In n8n 2.33.0 or later, runs of agents built with n8n’s Agents feature emit agent and LLM spans through the same exporter, including the model name and token counts. The AI Agent workflow node isn’t covered: it appears only as a node.execute span. Agent tracing is on by default once N8N_OTEL_ENABLED=true; set it explicitly to make the configuration clear:
By default, n8n agent tracing may record prompts, tool arguments, responses, and tool results. To reduce sensitive data capture:

​
Observe

  1. Run an n8n workflow.
  2. Open Phoenix and select the n8n project.
  3. Look for workflow.execute and node.execute spans. Runs of n8n Agents also appear as an agent span with LLM child spans if agent tracing is enabled.
If you configured OpenTelemetry from the n8n UI, use n8n’s Send test trace action to verify the exporter before running a workflow. n8n exports:
  • workflow.execute spans for workflow executions
  • node.execute spans for node executions
  • W3C trace context propagation for inbound webhooks and outbound HTTP requests
  • agent spans using OpenTelemetry GenAI semantic conventions when agent tracing is enabled

​
Troubleshooting

  • No traces in Phoenix. Confirm N8N_OTEL_ENABLED=true, the endpoint is the base Phoenix URL that n8n can reach, and the tracing path is /v1/traces. For an instance with authentication, confirm the authorization header carries a valid API key. n8n logs an error at startup if it can’t reach the endpoint.
  • Traces land in the default project. Confirm OTEL_RESOURCE_ATTRIBUTES includes openinference.project.name=n8n on every n8n process.
  • Only production executions appear. n8n exports production executions by default. Set N8N_OTEL_TRACES_PRODUCTION_ONLY=false to trace all executions, including manual test runs from the editor.
  • Node spans are missing. Confirm N8N_OTEL_TRACES_INCLUDE_NODE_SPANS=true.
  • Agent spans are missing. Confirm your n8n version supports agent tracing and N8N_AGENTS_TRACING_ENABLED=true.

​
Resources

n8n OpenTelemetry tracing

n8n OpenTelemetry environment variables