Skip to main content

OpenAI Agents SDK

Last updated on

Use this integration for OpenAI Agents SDK applications. Add the OpenInference OpenAI Agents instrumentation and point it at Harness. This is a code-based integration, because the OpenAI Agents SDK has no external tracing service to route from. The instrumentation captures the full workflow: function calls, tool uses, and LLM interactions, not just the LLM call.

Before you start, generate an ingestion token and connect a provider. Go to the AI Cost Management Quickstart for those steps. Replace <YOUR_TOKEN> below with that token and adjust the endpoint for your Harness cluster.

GenAI semantic conventions required

Cost is calculated from OpenTelemetry traces with GenAI semantic conventions. Go to the GenAI Span Attribute Reference to review the attributes CACM reads.

Instrument the OpenAI Agents SDK​

The OpenAI Agents SDK has built-in tracing. Use the OpenInference OpenAI Agents instrumentation to export traces to Harness.

Install the instrumentation library:

pip install openinference-instrumentation-openai-agents \
  opentelemetry-sdk opentelemetry-exporter-otlp

Configure the OpenTelemetry exporter once at application startup, then add the framework instrumentor:

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from openinference.instrumentation.openai_agents import OpenAIAgentsInstrumentor

provider = TracerProvider()
provider.add_span_processor(
    BatchSpanProcessor(
        OTLPSpanExporter(
            endpoint="https://app.harness.io/udp-ingest/otel/v1/traces",
            headers={"Authorization": "Bearer <YOUR_TOKEN>"},
        )
    )
)
trace.set_tracer_provider(provider)

OpenAIAgentsInstrumentor().instrument(tracer_provider=provider)

What this produces:

  • One trace per agent run.
  • Nested spans for function calls, tool uses, and LLM interactions.
  • GenAI semantic conventions on LLM spans.
  • Cost calculated from token counts.

Go to the OpenAI Agents SDK tracing docs to configure custom processors.

Reduce Trace Data Volume​

Large payloads and over-instrumentation inflate span volume and storage cost. To keep trace data manageable in high-traffic production:

  • Scope instrumentation to LLM calls: Instrument the model calls that carry cost, not every function in the application.
  • Sample a percentage of traces: Export a representative sample rather than every trace.

Verify Traces in Cost Explorer​

  1. Run the application so traces flow. They usually appear within a few minutes; allow up to about 20 minutes.
  2. Go to Cloud & AI Cost Management > Cost Explorer.
  3. Select the AI Traces view or group by Service Name, and find your service.
  4. Select a service row to open the Service Traces drawer and inspect the span waterfall.

Next Steps​