Glossary

What is AgentOps?

AgentOps is the operational discipline of running AI agents in production: observing, debugging, evaluating, and monitoring their behavior over time, by analogy to DevOps and MLOps. AgentOps is also the name of a company, agentops.ai, that builds observability tooling for agents; the two senses share a name and are worth keeping distinct.

As a practice, AgentOps is centered on observability and evaluation: instrumenting runs, catching regressions, measuring cost and quality as agents ship. It is the operations-facing half of running agents well.

Because the frame is operational and after-the-fact, it does not, on its own, cover the parts that decide what an agent may do in the first place: scoped permissions, durable memory, review gates that hold risky work before it lands, and a human intervention path. Those belong to a broader frame: one that treats agents less like a pipeline to monitor and more like a team to manage.

How it relates to agent management

AgentOps is one operational slice of the wider discipline of AI agent management, which spans the whole agent lifecycle: context and permissions and review, going well beyond observation.