AgentOps tool

Continuing with the series on LLMOps and agents, there are now tools available for monitoring and observing agent solutions. AgentOps caught my attention as I am trying to use CrewAI for building multi-agent systems, and they seem to have excellent integration with Crew.

The basic premise is: “Build your next agent with evals, observability, and replays.”

AgentOps provides a dashboard that displays a wealth of useful information about your agents, such as registering agent events and showing chat responses for these events.

The way to start your dashboard is basically:

pip install agentops

And then

import agentops
agentops.init(<INSERT YOUR API KEY HERE>)

you have to make your AgentOps key that you can find here:

https://app.agentops.ai/settings/projects

After that you have a dashboard deployed in an URL similar to:

https://app.agentops.ai/drilldown

After that its a matter of using, agentops classes and decorators, for example:

import openai
import agentops

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

# (record specific functions)
@agentops.record_action('sample function being record')
def…