Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Arize-ai/phoenix/llms.txt

Use this file to discover all available pages before exploring further.

Phoenix provides auto-instrumentation for the Anthropic Python SDK and Anthropic TypeScript SDK, allowing you to trace all calls to Claude models. Anthropic’s Claude models are state-of-the-art LLMs with extended context windows and strong performance on reasoning tasks. Phoenix makes it easy to observe Claude’s behavior in production.

​
Installation

pip install openinference-instrumentation-anthropic anthropic

​
Setup

1

Set your Anthropic API key

export ANTHROPIC_API_KEY=sk-ant-...
2

Register Phoenix tracer

from phoenix.otel import register

# Configure the Phoenix tracer
tracer_provider = register(
  project_name="my-llm-app",  # Default is 'default'
  auto_instrument=True  # Auto-instrument based on installed packages
)
3

Use Anthropic as normal

import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1000,
    temperature=0,
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "text",
                    "text": "Why is the ocean salty?"
                }
            ]
        }
    ]
)
print(message.content)

​
What Gets Traced

The Anthropic instrumentation automatically captures:
  • Messages: Model, system prompt, messages, parameters
  • Streaming: Real-time token streaming
  • Tool Use: Tool definitions, tool calls, results
  • Token Usage: Input and output token counts
  • Errors: API errors, rate limits, validation errors

​
Advanced Examples

​
Streaming Responses

import anthropic

client = anthropic.Anthropic()

with client.messages.stream(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Write a short poem about AI observability"}
    ],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

# Phoenix captures the full streamed response

​
Tool Use (Function Calling)

import anthropic
import json

client = anthropic.Anthropic()

# Define tools
tools = [
    {
        "name": "get_weather",
        "description": "Get the current weather in a given location",
        "input_schema": {
            "type": "object",
            "properties": {
                "location": {
                    "type": "string",
                    "description": "The city and state, e.g. San Francisco, CA"
                },
                "unit": {
                    "type": "string",
                    "enum": ["celsius", "fahrenheit"],
                    "description": "The unit of temperature"
                }
            },
            "required": ["location"]
        }
    }
]

message = client.messages.create(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "What's the weather like in San Francisco?"}]
)

# Phoenix captures tool definitions and calls
if message.stop_reason == "tool_use":
    for content in message.content:
        if content.type == "tool_use":
            print(f"Tool: {content.name}")
            print(f"Input: {content.input}")

​
System Prompts and Multi-turn Conversations

import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1024,
    system="You are a helpful AI assistant specialized in Python programming.",
    messages=[
        {"role": "user", "content": "How do I read a CSV file?"},
        {"role": "assistant", "content": "You can use the pandas library..."},
        {"role": "user", "content": "What if I don't have pandas?"}
    ]
)

print(message.content[0].text)

​
Observability in Phoenix

Once instrumented, you can:
  • View all Claude API calls with full message history
  • Monitor token usage for cost tracking
  • Track response latency and streaming performance
  • Inspect tool use patterns and tool call sequences
  • Debug system prompts and conversation context
  • Analyze error patterns and rate limiting

​
Resources

Example Messages

Basic message example

Tool Calling Example

Multiple tool calls

OpenInference Package

View source code

Anthropic Documentation

Official API docs

Build docs developers (and LLMs) love

Get started for freeTalk to us