What is Tessary's tool_error classifier?

tool_error is one of Tessary’s built-in classifiers: it watches the rate at which a tool’s calls fail, not any single call. A call counts as failed only on structural evidence, an error status on the span, a recorded exception, or a result that declares itself an error, never on output text that merely mentions the word error.

Each tool gets its own alert threshold, set from that tool’s own normal failure rate and a false-alarm budget, rather than one fixed cutoff shared across every tool. That way a chatty, error-prone integration and a normally reliable API aren’t judged against the same bar. Evidence accumulates call by call, so a tool failing at its usual rate stays quiet, and only a sustained shift fires.

A detection is one finding per tool: when the rate moved, which failure patterns drove it, and which call sites the traffic came through. Triage confirms a real deviation before it becomes a case anyone has to act on.