AI agent observability is the ability to see what an AI agent did on every run — inputs, retrieved evidence, model calls, outputs, cost and errors — so it can be monitored, debugged and audited.
Also called: AgentOps, LLM observability, AI agent monitoring, LLM tracing
Traditional monitoring tells you whether a service is up. An AI agent can be up and still wrong. Observability for agents therefore records the content of each run: what the agent was asked, which documents it retrieved, which model it called with what prompt, what each step returned, how long it took and what it cost.
Together with tests that run before release and alerts when runs fail, this is the operational discipline often called AgentOps. It turns a vague complaint about an answer into a specific step to fix, and gives auditors a record of what the system actually did.
This site needs JavaScript for the full experience. Every page above is reachable without it.