Jev is TypeSafe AI's System One API for discovering available models and evaluating structured state with Noul, Choice, and Score questions.
Every capability is a discrete, logged action the agent calls by name — scoped to what you authorize and recorded in the run trace.