English

ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox

Artificial Intelligence 2026-05-21 v2 Software Engineering

Abstract

Current LLM agents are proficient at calling isolated APIs but struggle with the "last mile" of commercial software automation. In real-world scenarios, tools are not independent; they are atomic, interdependent, and prone to environmental noise. We introduce ComplexMCP\textbf{ComplexMCP}, a benchmark designed to evaluate agents in these rigorous conditions. Built on the Model Context Protocol (MCP), ComplexMCP\textbf{ComplexMCP} provides over 300 meticulously tested tools derived from 7 stateful sandboxes, ranging from office suites to financial systems. Unlike existing datasets, our benchmark utilizes a seed-driven architecture to simulate dynamic environment states and unpredictable API failures, ensuring a deterministic yet diverse evaluation. We evaluate various LLMs across full-context and RAG paradigms, revealing a stark performance gap: even top-tier models fail to exceed a 60% success rate, far trailing human performance 90%. Granular trajectory analysis identifies three fundamental bottlenecks: (1) tool retrieval saturation\textbf{tool retrieval saturation} as action spaces scale; (2) over-confidence\textbf{over-confidence}, where agents skip essential environment verifications; and (3) strategic defeatism\textbf{strategic defeatism}, a tendency to rationalize failure rather than pursuing recovery. These findings underscore the insufficiency of current agents for interdependent workflows, positioning ComplexMCP\textbf{ComplexMCP} as a critical testbed for the next generation of resilient autonomous systems.

Keywords

Cite

@article{arxiv.2605.10787,
  title  = {ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox},
  author = {Yuanyang Li and Xue Yang and Longyue Wang and Weihua Luo and Hongyang Chen},
  journal= {arXiv preprint arXiv:2605.10787},
  year   = {2026}
}
R2 v1 2026-07-22T07:04:53.170Z