OpenInference OpenAI Agents Instrumentation

Python auto-instrumentation library for OpenAI Agents python SDK.

The traces emitted by this instrumentation are fully OpenTelemetry compatible and can be sent to an OpenTelemetry collector for viewing, such as Arize Phoenix or Arize AX.

Compatibility

`openinference-instrumentation-openai-agents`	`openai-agents`	Python
`>=2.0`	`>=0.11.0`	`>=3.10, <3.15`
`>=1.4.1, <2.0`	`>=0.2.6`	`>=3.10, <3.15`
`>=1.2.0, <1.4.1`	`>=0.2.6`	`>=3.9, <3.14`
`>=1.0.0, <1.2.0`	`>=0.1.0`	`>=3.9, <3.14`
`<1.0.0`	`>=0.0.3`	`>=3.9, <3.14`

Instrumentor >=2.0 requires openai-agents>=0.11.0. Three things changed below that floor which the instrumentor no longer accommodates: the run internals moved out of agents._run_impl into agents.run_internal in 0.8.0, openai-agents moved from openai<2 to openai>=2.9 in the same release, and tool namespaces (agents.tool_namespace) arrived in 0.11.0. Tools grouped by a namespace report tool.description and tool.parameters only on instrumentor >=2.0.

Installation

pip install openinference-instrumentation-openai-agents

Quickstart

In this example we will instrument a small program that uses OpenAI and observe the traces via arize-phoenix.

Install packages.

pip install openinference-instrumentation-openai-agents arize-phoenix opentelemetry-sdk opentelemetry-exporter-otlp

Start the phoenix server so that it is ready to collect traces. The Phoenix server runs entirely on your machine and does not send data over the internet.

phoenix serve

In a python file, set up the OpenAIAgentsInstrumentor and configure the tracer to send traces to Phoenix.

from agents import Agent, Runner
from openinference.instrumentation.openai_agents import OpenAIAgentsInstrumentor
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk import trace as trace_sdk
from opentelemetry.sdk.trace.export import ConsoleSpanExporter, SimpleSpanProcessor

endpoint = "http://127.0.0.1:6006/v1/traces"
tracer_provider = trace_sdk.TracerProvider()
tracer_provider.add_span_processor(SimpleSpanProcessor(OTLPSpanExporter(endpoint)))
# Optionally, you can also print the spans to the console.
tracer_provider.add_span_processor(SimpleSpanProcessor(ConsoleSpanExporter()))

OpenAIAgentsInstrumentor().instrument(tracer_provider=tracer_provider)


agent = Agent(name="Assistant", instructions="You are a helpful assistant")
result = Runner.run_sync(agent, "Write a haiku about recursion in programming.")
print(result.final_output)

Since we are using OpenAI, we must set the OPENAI_API_KEY environment variable to authenticate with the OpenAI API.

export OPENAI_API_KEY=your-api-key

Now simply run the python file and observe the traces in Phoenix.

python your_file.py

Realtime audio

OpenAIAgentsInstrumentor().instrument(...) also traces agents.realtime.RealtimeSession (the OpenAI Agents SDK's voice/audio runtime) when the realtime extras are installed. No additional setup is required — instrument(...) applies the realtime patches whenever agents.realtime is importable.

For each turn the instrumentor produces this span tree:

AUDIO   "conversation.turn"     ← parent; aggregated input/output transcripts, llm.model_name, llm.invocation_parameters
├─ USER  "user"                 ← input.audio.url (WAV data URI), input.audio.transcript, or input.value for text input
├─ LLM   "assistant"            ← output.audio.url, output.audio.transcript, token counts, time_to_first_token_ms
│  └─ TOOL "<tool_name>"        ← one per function call within the turn
└─ ...                          ← additional USER / LLM siblings for split input or tool round-trips

A runnable mic/speaker example with two function tools lives at examples/realtime_with_tools.py.

Audio redaction

The realtime instrumentor recognizes three environment variables for redacting captured audio:

OPENINFERENCE_HIDE_INPUT_AUDIO — when truthy (1 / true / yes / on), drops input.audio.url, input.audio.mime_type, and input.audio.transcript from USER spans. Default: false.
OPENINFERENCE_HIDE_OUTPUT_AUDIO — same shape, drops the output.audio.* attributes from LLM spans. Default: false.
OPENINFERENCE_BASE64_AUDIO_MAX_LENGTH — caps the base64 payload length of audio data: URIs. The data:audio/wav;base64, prefix is always preserved. Default: 32000.

TraceConfig(hide_inputs=True) and TraceConfig(hide_outputs=True) also cascade to the corresponding audio attributes.

More Info

Metadata

Release files for openinference-instrumentation-openai-agents 2.2.2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for openinference-instrumentation-openai-agents 2.2.2
File	Size	Uploaded
openinference_instrumentation_openai_agents-2.2.2.tar.gz	32.1 kB	Sep 9, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for openinference-instrumentation-openai-agents 2.2.2
File	Interpreter	ABI	Platform	Reset
openinference_instrumentation_openai_agents-2.2.2-py3-none-any.whl 35.2 kB Sep 9, 2026	Python 3	none	any	Details

Total release size: 67.4 kB

Release files / openinference_instrumentation_openai_agents-2.2.2.tar.gz

Download URL	openinference_instrumentation_openai_agents-2.2.2.tar.gz
Size	32.1 kB
Tags	Source
SHA-256 checksum How to use checksums	`0242b27659972ab3d8c4953ebb9c722ac03a091f3eb936fa73a69a6a5ffd983e`
BLAKE2b-256 checksum How to use checksums	`b8c061175accb5fdad137d36cd973e0134a40050993d3c92de47ff4022a86acf`
Upload date	Sep 9, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/Arize-ai/openinference
Publishing commit: github.com/Arize-ai/openinference/tree/8abaf899cb9a039284663b2f65cb63738d26b1ed

Workflow

Publishing configuration: github.com/Arize-ai/openinference/blob/8abaf899cb9a039284663b2f65cb63738d26b1ed/.github/workflows/publish.yaml
Publishing logs: github.com/Arize-ai/openinference/actions/runs/34322596540/attempts/1

Artifact

Subject: openinference_instrumentation_openai_agents-2.2.2.tar.gz
SHA-256 checksum: 0242b27659972ab3d8c4953ebb9c722ac03a091f3eb936fa73a69a6a5ffd983e Verifying attestations

Release files / openinference_instrumentation_openai_agents-2.2.2-py3-none-any.whl

Download URL	openinference_instrumentation_openai_agents-2.2.2-py3-none-any.whl
Size	35.2 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`eba1554fe98825915035bbfc61b978ef753251b248896e28160366b1449536af`
BLAKE2b-256 checksum How to use checksums	`f5d2edde4339d6be77aff2c91dce7af5dc002bfc474282082356d3196953265e`
Upload date	Sep 9, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/Arize-ai/openinference
Publishing commit: github.com/Arize-ai/openinference/tree/8abaf899cb9a039284663b2f65cb63738d26b1ed

Workflow

Publishing configuration: github.com/Arize-ai/openinference/blob/8abaf899cb9a039284663b2f65cb63738d26b1ed/.github/workflows/publish.yaml
Publishing logs: github.com/Arize-ai/openinference/actions/runs/34322596540/attempts/1

Artifact

Subject: openinference_instrumentation_openai_agents-2.2.2-py3-none-any.whl
SHA-256 checksum: eba1554fe98825915035bbfc61b978ef753251b248896e28160366b1449536af Verifying attestations