Braintrust

Braintrust traces what an AI application actually did in production — prompts, tool calls, outputs — and lets teams run evaluations against datasets to catch quality regressions before they ship. It's part of the newer wave of LLM ops tooling built around the reality that AI features fail in ways traditional software testing doesn't catch.