Enables running LLM evaluations, experiments, and custom evaluators through a standardized MCP interface.