What is a trace?

A trace is the stored record of one agent turn: one thing a user asked for, and everything the agent did to answer it, from the first LLM call to the last tool result. It sits in the middle of Tessary’s three fixed levels, below the session that holds a user’s whole run of turns and above the spans that are the individual steps inside a turn, an LLM call, a tool call, or a sub-agent invocation. A trace is keyed on the ids your own instrumentation already sends; nothing gets invented on Tessary’s side, so it ties directly back to the request that produced it. Classifiers, cases, and root cause analysis all read this same stored trace, not a summary of it, which is why a finding always comes with the actual evidence behind it open and waiting.