Hosted decision APIs and seven MCP judgment/grading tools for developers building classification, routing and rubric-based evaluation. Submitted by the Jev Agent team; independent, not affiliated with TypeSafe. Returns bounded judgments, not generated prose or code