# What's the difference between a custom classifier and a built-in one?

A custom classifier differs from a built-in only in what it's tuned to: built-ins run the same generic check across every customer, because they catch failure modes every agent shares; a custom classifier is fine-tuned on your own agent's traffic to catch failures that are yours alone. Built-in classifiers cover things like frustration, groundedness, secret_leak, and malformed_output. A custom classifier covers the policy your agent has to follow, the mistake it keeps making, the behavior your team already knows to watch for that a generic check was never going to see.

Mechanically the two are identical. A custom classifier runs on every trace at the same negligible cost as a built-in, scores against a confidence threshold, and files a finding when it crosses that threshold. It doesn't replace the built-ins, it runs alongside them, adding coverage the generic set can't provide by design.

---

Source: https://tessary.ai/answers/tessary-custom-classifiers/difference-between-custom-and-built-in-classifier
More on Tessary custom classifiers: https://tessary.ai/answers/tessary-custom-classifiers
From Tessary, agent reliability for AI agents in production: https://tessary.ai