How to beginner · 3 min read

How to track agent costs with AgentOps

Q: How to track agent costs with AgentOps

Use the agentops Python package to initialize cost tracking by calling agentops.init() with your API key. Wrap your agent calls within sessions using agentops.start_session() and agentops.end_session() to automatically monitor and log usage and costs.

Quick answer

Use the agentops Python package to initialize cost tracking by calling agentops.init() with your API key. Wrap your agent calls within sessions using agentops.start_session() and agentops.end_session() to automatically monitor and log usage and costs.

PREREQUISITES

Python 3.8+
AgentOps API key
pip install agentops
OpenAI API key (if using OpenAI models)

Setup

Install the agentops package and set your environment variables for AGENTOPS_API_KEY and your AI provider API key (e.g., OPENAI_API_KEY).

Run pip install agentops
Export environment variables in your shell or set in your IDE:

export AGENTOPS_API_KEY=<your_agentops_key>
export OPENAI_API_KEY=<your_openai_key>

bash

pip install agentops

output

Collecting agentops
  Downloading agentops-1.0.0-py3-none-any.whl (10 kB)
Installing collected packages: agentops
Successfully installed agentops-1.0.0

Step by step

Initialize AgentOps in your Python script, start a session to track an agent's execution, and end the session to finalize cost logging.

python

import os
import agentops
from openai import OpenAI

# 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 session to track costs and usage
session = agentops.start_session(tags=["example-agent"])

try:
    # Make a chat completion call
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": "Hello, how much does this cost?"}]
    )
    print("Response:", response.choices[0].message.content)
finally:
    # End the session and log costs
    agentops.end_session("Completed successfully")

output

Response: The cost depends on the tokens used, but this call is tracked by AgentOps.
Session ended: Completed successfully

Common variations

You can use agentops with async code by awaiting agentops.start_session() and agentops.end_session(). AgentOps also auto-instruments OpenAI calls if agentops.init() is called before client usage. You can track multiple sessions or tag sessions for grouping costs.

python

import asyncio
import os
import agentops
from openai import OpenAI

async def main():
    agentops.init(api_key=os.environ["AGENTOPS_API_KEY"])
    client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

    session = await agentops.start_session(tags=["async-agent"])
    try:
        response = await client.chat.completions.create(
            model="gpt-4o-mini",
            messages=[{"role": "user", "content": "Async cost tracking example."}]
        )
        print("Async response:", response.choices[0].message.content)
    finally:
        await agentops.end_session("Async session complete")

asyncio.run(main())

output

Async response: This is an example of async cost tracking with AgentOps.
Session ended: Async session complete

Troubleshooting

If you see no cost data in your AgentOps dashboard, ensure agentops.init() is called before any AI client usage.
Verify your AGENTOPS_API_KEY is correct and has permissions.
For missing session logs, always call agentops.end_session() to flush data.

✅

Key Takeaways

Initialize AgentOps early in your app with agentops.init() to enable automatic cost tracking.
Wrap your AI calls within agentops.start_session() and agentops.end_session() to monitor usage and expenses.
AgentOps supports both synchronous and asynchronous Python workflows for flexible integration.

Verified 2026-04 · gpt-4o-mini

Verify ↗