What is Tessary's cost_drift classifier?

cost_drift is one of Tessary’s built-in classifiers. It watches how much a call site spends per turn and compares that against the call site’s own recent history, not a fixed dollar figure.

It catches the ways an agent gets quietly more expensive: a prompt that grew, retries that multiplied, caching that stopped working. None of that throws an error or fails a test, so nothing else in the stack notices until the bill does.

Comparisons run against both the call site’s recent normal and a reference pinned at the last deploy, so a sudden jump and a slow creep both register, and a drop in cost shows up just as clearly as a rise. A detection is a finding, not an alert straight to a person; triage decides whether it’s a real deviation before it becomes a case.

keep reading

More on this.

Can cost_drift catch a cost that creeps up slowly, not just a sudden spike?Yes. It checks a rolling recent baseline that catches sudden jumps and one pinned at the last deploy that catches the creep the rolling one absorbs.answer →Does cost_drift work for a scheduled agent that has no user sessions?Yes. cost_drift tracks dollars per turn at the call site, not per user session, so a scheduled job's runs build a baseline the same way.answer →Is cost_drift turned on by default?Yes. cost_drift is one of three classifiers Tessary calls deterministic and turns on by default, alongside duration_drift and tool_error.answer →Everything on tessary cost driftThe definition, and every question answered under tessary cost drift.concept →

Send us the traces you already emit.

book a demo