How to beginner · 3 min read

How to use AgentOps with OpenAI

Q: How to use AgentOps with OpenAI

Use agentops.init() to automatically patch the OpenAI client for observability. Start and end manual sessions with agentops.start_session() and agentops.end_session() to track AI agent interactions seamlessly.

Quick answer

Use agentops.init() to automatically patch the OpenAI client for observability. Start and end manual sessions with agentops.start_session() and agentops.end_session() to track AI agent interactions seamlessly.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
AgentOps API key
pip install openai>=1.0 agentops

Setup

Install the required packages and set environment variables for OPENAI_API_KEY and AGENTOPS_API_KEY. This enables authentication for both OpenAI and AgentOps services.

bash

pip install openai agentops

output

Collecting openai
Collecting agentops
Successfully installed openai-1.x.x agentops-x.x.x

Step by step

Initialize AgentOps to auto-instrument OpenAI calls, then create an OpenAI client and make a chat completion request. Use manual session management to track specific agent interactions.

python

import os
from openai import OpenAI
import agentops

# Initialize AgentOps with your API key
agentops.init(api_key=os.environ["AGENTOPS_API_KEY"])

# Create OpenAI client
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

# Start a manual session (optional)
session = agentops.start_session(tags=["example-agent"])

# Make a chat completion request
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello, AgentOps!"}]
)

print("Response:", response.choices[0].message.content)

# End the session
agentops.end_session("Success")

output

Response: Hello, AgentOps!

Common variations

Use agentops.start_session() and agentops.end_session() to manually control session boundaries for complex workflows.
AgentOps auto-instruments all OpenAI client calls after agentops.init(), so no code changes are needed for basic usage.
Combine with async OpenAI calls by initializing AgentOps before async usage.

Troubleshooting

If you see no data in AgentOps dashboard, verify AGENTOPS_API_KEY is set correctly and agentops.init() is called before any OpenAI client usage.
Ensure your OpenAI API key is valid and environment variables are loaded properly.
For network issues, check firewall or proxy settings that might block telemetry data.

✅

Key Takeaways

Call agentops.init() early to auto-instrument OpenAI client calls for observability.
Use manual sessions with agentops.start_session() and agentops.end_session() to track agent workflows explicitly.
Set environment variables OPENAI_API_KEY and AGENTOPS_API_KEY securely before running your code.

Verified 2026-04 · gpt-4o-mini

Verify ↗