Braintrust

Evaluation and observability for AI

Braintrust is an AI observability platform that connects evaluations and production monitoring in one workflow, giving builders visibility into how their AI actually behaves once deployed and the tooling to improve it. Keeping evals and observability together is the substantive design choice: teams that separate them end up measuring one thing in testing and watching another in production, with no path between the two. Its users are software engineers, and its early customers already include Stripe, Notion, Instacart and Airtable.