agentops

What it is

AgentOps-AI/agentops is a Python SDK for AI agent monitoring, LLM cost tracking, benchmarking, and more. It integrates with major LLMs and agent frameworks including CrewAI, Agno, OpenAI Agents SDK, Langchain, Autogen, AG2, and CamelAI.

How it works

The repository exposes an SDK that offers decorators for session, agent, operation, task, and workflow to annotate code and collect observability data. It provides integrations with multiple frameworks (OpenAI Agents SDK, CrewAI, AG2, Langchain, Camel, Cohere, Anthropic, etc.). It supports self-hosting the AgentOps app and provides a quick start via pip install agentops. Example usage shows initializing the client, starting a session, and ending sessions to capture analytics.

Getting started

pip install agentops

import agentops

# Beginning of your program (i.e. main.py, __init__.py)
agentops.init( < INSERT YOUR API KEY HERE >)

...

# End of program
agentops.end_session('Success')

All sessions can be viewed on the AgentOps dashboard: https://app.agentops.ai?ref=gh

First class developer experience examples include:

from agentops.sdk.decorators import session

@session
def my_workflow():
    return result

from agentops.sdk.decorators import agent

@agent
class MyAgent:
    def __init__(self, name):
        self.name = name

from agentops.sdk.decorators import operation, task

@operation
def process_data(data):
    return result

from agentops.sdk.decorators import workflow

@workflow
def my_workflow(data):
    return result

Integrations include:

• OpenAI Agents SDK (Python: pip install openai-agents; TypeScript: npm install agentops @openai/agents)
• CrewAI (pip install 'crewai[agentops]')
• AG2 (AutoGen) observability
• Camel AI (installation and example provided in docs)
• LangChain (optional dependency: pip install agentops[langchain])
• Cohere (pip install cohere; example usage)
• Anthropic (pip install anthropic; example usage)

Usage notes indicate that AgentOps can be initialized and used to create session/agent/operation/workflow spans, and that integrations enable observation across various frameworks.

Recent releases

• 0.4.21 (2025-08-29): add back lib by @areibman; updated gitignore; new screensho
• 0.4.20 (2025-08-15): Add video embeds to docs; AgentOps OSS release
• 0.4.19 (2025-08-01): docs: update examples from notebooks; Create openai responses api test
• 0.4.18 (2025-07-17): [tests] coverage dopamine; fix mcp docs
• 0.4.17 (2025-07-01): Add separate docs for OpenAI Agents SDKs; Update session management documentation

Traction

• Stars: 5750

Behind the repo

• The project is open source under the MIT license and has a MIT badge in the README.

Caveats

• License: MIT
• Created: 2023-08-15
• Last push: 2026-06-25
• Open issues: 177
• Forks: 613
• Language: Python
• Topics include agent, llm, langchain, openai, anthropic, etc.