Skip to content

Langflow

See each Langflow run as a trace (the full journey of one flow run, made of one or more spans) in Logfire. A span is one unit of work: a single operation, with a name, a start, and a duration. The trace includes instrumented model and tool activity.

Langflow sends traces using the OpenTelemetry Protocol (OTLP), the standard wire format Logfire uses to receive telemetry. This route requires an administrator who can set the Langflow service environment; it is not available in a hosted workspace that does not expose runtime settings.

Configure Langflow

Create a project write token, then set:

TRACELOOP_API_KEY=<your-write-token>
TRACELOOP_BASE_URL=https://logfire-us.pydantic.dev
OTEL_EXPORTER_OTLP_INSECURE=false

For an EU project, use https://logfire-eu.pydantic.dev. Langflow requires TRACELOOP_API_KEY to initialize its tracing integration. The instrumentation sends that value as a Bearer authorization token, and appends /v1/traces to TRACELOOP_BASE_URL.

Do not also set TRACELOOP_HEADERS: when present, it takes precedence over TRACELOOP_API_KEY in the underlying instrumentation library.

Restart Langflow after applying the environment.

Verify

Run a flow, then open the Live view. You should see a root Langflow span with child spans for instrumented components.

Troubleshooting

SymptomCause and fix
No traces arriveConfirm TRACELOOP_API_KEY is non-empty and Langflow can reach the regional URL.
Requests go to the wrong pathSupply the Logfire base URL without /v1/traces; the instrumentation appends the path.
Authentication failsReplace TRACELOOP_API_KEY with a valid write token and remove any conflicting TRACELOOP_HEADERS.

See Langflow’s Traceloop integration and Traceloop configuration for the vendor settings used by this route.