Braintrust

Braintrust is an AI evaluation and observability platform that helps teams measure and improve the quality of LLM apps and agents. It traces every call in production, runs automated and LLM-as-judge evals on datasets, compares prompts and models in a playground, and catches regressions in CI. For engineering and product teams building and iterating on AI products.

• Production tracing
• Automated and LLM-judge evals
• Prompt and model comparison