Skip to main content
Looking for TypeScript? See the TypeScript guide.
PyPI Version This module provides OpenInference instrumentation for Anthropic’s Claude Agent SDK, automatically capturing AGENT and TOOL spans that follow OpenInference semantic conventions.

​
Install

pip install openinference-instrumentation-claude-agent-sdk claude-agent-sdk arize-phoenix-otel

​
Setup

Use the register function to connect your application to Phoenix: 
from phoenix.otel import register

tracer_provider = register(
  project_name="claude-agent-sdk",
  auto_instrument=True,
)

​
Run Claude Agent SDK

A simple Claude Agent SDK application that is now instrumented: 
import asyncio
from claude_agent_sdk import query, ClaudeAgentOptions, AssistantMessage, TextBlock


async def main():
    async for message in query(
        prompt="What files are in the current directory?",
        options=ClaudeAgentOptions(allowed_tools=["Bash", "Glob"]),
    ):
        if isinstance(message, AssistantMessage):
            for block in message.content:
                if isinstance(block, TextBlock):
                    print(block.text)


asyncio.run(main())

​
Observe

With instrumentation enabled, you will see the following in Phoenix:
  • AGENT spans wrapping the full query() call, capturing the prompt and final response
  • TOOL spans for each tool invocation made by the agent during execution (e.g., Bash commands, file reads)
These spans follow OpenInference semantic conventions, making them fully compatible with Phoenix’s trace visualization and evaluation features.

​
Privacy Configuration

If you need to hide sensitive data, use the instrumentor directly instead of auto_instrument: 
from phoenix.otel import register
from openinference.instrumentation.claude_agent_sdk import ClaudeAgentSDKInstrumentor

tracer_provider = register(project_name="claude-agent-sdk")

ClaudeAgentSDKInstrumentor().instrument(
    tracer_provider=tracer_provider,
    hide_inputs=True,
    hide_outputs=True,
)

​
Resources