← tools · evals & testing
Braintrust
Evals, observability, and a prompt playground for shipping AI features that don't quietly regress.
pricing
free tier; usage-based
where it wins
When 'did the model get worse this week' needs a dashboard instead of a vibe.
watch out
Usage pricing scales with the team — check the math before it becomes the default.
no affiliate program on file — listed because it belongs here.