Ingest traces, score gold sets, export data. Hosted at eval.phthos.ai or self-host with pip. Eval does not use Gateway or Task AI credits.
Push traces from Task AI or your own stack with a pk_ key.
Score against labeled examples. Export anytime.
pip install phthos-eval. Self-host has no volume cap.
Pro uses Phthos Gateway as a judge — billed on Eval, not console.phthos.ai.
Volume caps, not Phthos credits. Checkout at eval.phthos.ai/billing.
$49/mo
Usage and plan live at eval.phthos.ai/plan and /usage.