Create custom agent evaluation datasets with mocked APIs, databases, and custom file systems, and run them in a fully managed sandboxed environment uniquely configured to match your production settings, with full tracing and actionable insights.