MS-Eval: Benchmarking Sub-Agent Delegation and Orchestration in MCP-Intensive Working Scenarios

Explain Like I'm Normal

Researchers introduced MS-Eval to measure how well 'Main-Agents' delegate complex tasks to specialized 'Sub-Agents' using the Model Context Protocol (MCP). This shift from sequential tool-calling to parallel sub-agent orchestration aims to solve context bloat and improve efficiency in complex workflows. The benchmark evaluates both the quality of the delegation and the final orchestration of results across frontier models.