OpenTelemetry or recorder
OpenAI Agents SDK
The Agents SDK has its own tracing. Bridge it to OpenTelemetry with an instrumentation library such as OpenInference and import the traces, or record each run's usage directly.
- Either add an OpenTelemetry instrumentation for the Agents SDK and follow the OpenTelemetry panel,
- or call the recorder after each
Runner.runwith the run's usage (below).
Record a run directly
from harness_recorder import Recorder
rec = Recorder("support-triage", "2.1.0")
run = rec.start_run(ticket_id)
with run.agent("draft_reply", prompt_ref="prompts/drafter@6", model="openai/gpt-5.6-terra") as step:
result = await Runner.run(agent, ticket_text)
u = result.context_wrapper.usage
step.model_call(input_tokens=u.input_tokens, output_tokens=u.output_tokens)
run.finish("completed")Where the log lands
In your collector's trace file, or wherever your service logs go.
How it reaches the harness
You copy the file or download the logs with your own access.