How to beginner · 3 min read

AgentOps multi-agent tracking

Q: AgentOps multi-agent tracking

Use the agentops Python SDK to initialize tracking with agentops.init() and create separate sessions for each agent via agentops.start_session(). Track multiple agents concurrently by managing distinct session IDs and ending each session with agentops.end_session().

Quick answer

Use the agentops Python SDK to initialize tracking with agentops.init() and create separate sessions for each agent via agentops.start_session(). Track multiple agents concurrently by managing distinct session IDs and ending each session with agentops.end_session().

PREREQUISITES

Python 3.8+
AgentOps API key
pip install agentops

Setup

Install the agentops package and set your API key as an environment variable for secure authentication.

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, start multiple agent sessions, log actions, and end sessions to track multi-agent workflows.

python

import os
import agentops

# Set your API key in environment variable AGENTOPS_API_KEY
api_key = os.environ["AGENTOPS_API_KEY"]

# Initialize AgentOps tracking
agentops.init(api_key=api_key)

# Start session for Agent 1
session1 = agentops.start_session(tags=["agent1"])
print(f"Started session for Agent 1: {session1}")

# Start session for Agent 2
session2 = agentops.start_session(tags=["agent2"])
print(f"Started session for Agent 2: {session2}")

# Simulate some agent activity
print("Agent 1 processing task...")
# ... your agent 1 logic here

print("Agent 2 processing task...")
# ... your agent 2 logic here

# End sessions with status
agentops.end_session(session1, "Success")
agentops.end_session(session2, "Success")

print("Both agent sessions ended successfully.")

output

Started session for Agent 1: 123e4567-e89b-12d3-a456-426614174000
Started session for Agent 2: 123e4567-e89b-12d3-a456-426614174001
Agent 1 processing task...
Agent 2 processing task...
Both agent sessions ended successfully.

Common variations

You can use asynchronous tracking with agentops in async environments or integrate automatic OpenAI client patching for seamless multi-agent observability.

python

import asyncio
import agentops

async def run_agents():
    agentops.init(api_key=os.environ["AGENTOPS_API_KEY"])

    session1 = agentops.start_session(tags=["agent1"])
    session2 = agentops.start_session(tags=["agent2"])

    print(f"Async sessions started: {session1}, {session2}")

    # Simulate async agent tasks
    await asyncio.sleep(1)

    agentops.end_session(session1, "Success")
    agentops.end_session(session2, "Success")

    print("Async agent sessions ended.")

asyncio.run(run_agents())

output

Async sessions started: 123e4567-e89b-12d3-a456-426614174002, 123e4567-e89b-12d3-a456-426614174003
Async agent sessions ended.

Troubleshooting

If you see Missing API key errors, ensure AGENTOPS_API_KEY is set in your environment.
Sessions not tracked? Confirm agentops.init() is called before starting sessions.
For concurrency issues, manage session IDs carefully to avoid overlap.

✅

Key Takeaways

Initialize AgentOps once per application run with agentops.init() using your API key.
Create separate sessions for each agent with agentops.start_session() to track multi-agent workflows distinctly.
Always end sessions with agentops.end_session() to finalize tracking and report status.
Use async patterns for concurrent agent tracking in asynchronous applications.
Check environment variables and initialization order to avoid common tracking errors.

Verified 2026-04

Verify ↗