Human Tool: An MCP-Style Framework for Human-Agent Collaboration

TL;DR

The paper tackles the problem that human-led orchestration in AI-rich tasks can degrade performance when AI outperforms humans. It introduces Human Tool, an MCP-style interface that treats humans as callable tools with defined capabilities, information, and authority, enabling AI to orchestrate proactive workflows while selectively invoking human input. The authors operationalize this framework and validate it through controlled experiments on decision-making and creative tasks, showing improved task performance, reduced workload, and more balanced collaboration dynamics compared to an AI Tool baseline. The results demonstrate that AI-led coordination, when coupled with structured human input at strategic points, can preserve human agency, enhance robustness, and scale collaboration effectively. This work suggests a broader design direction where MCP-style abstractions extend to human contributors, offering practical implications for human-centered AI design and AI-driven workflow management.

Abstract

Human-AI collaboration faces growing challenges as AI systems increasingly outperform humans on complex tasks, while humans remain responsible for orchestration, validation, and decision oversight. To address this imbalance, we introduce Human Tool, an MCP-style interface abstraction, building on recent Model Context Protocol designs, that exposes humans as callable tools within AI-led, proactive workflows. Here, "tool" denotes a coordination abstraction, not a reduction of human authority or responsibility. Building on LLM-based agent architectures, we operationalize Human Tool by modeling human contributions through structured tool schemas of capabilities, information, and authority. These schemas enable agents to dynamically invoke human input based on relative strengths and reintegrate it through efficient, natural interaction protocols. We validate the framework through controlled studies in both decision-making and creative tasks, demonstrating improved task performance, reduced human workload, and more balanced collaboration dynamics compared to baseline systems. Finally, we discuss implications for human-centered AI design, highlighting how MCP-style human tools enable strong AI leadership while amplifying uniquely human strengths.

Figures (17)

• Figure 1: Contrasting paradigms of human-AI collaboration: Human Tool versus AI Tool. In the Human Tool paradigm, humans are exposed to the agent as MCP-style callable interfaces rather than workflow leaders.
• Figure 2: An integrated framework for representing Human Tool within AI-managed workflows. It organizes three dimensions: How to define Human Tool, when to call them, and how to communicate effectively.
• Figure 3: MCP-style orchestration architecture of the Human Tool framework, illustrating tool schema definition, invocation decisions, and tool call-response integration.
• Figure 4: Comparison of Accuracy and Mental Effort Across Tasks (Travel Planning vs. Story Writing) Between Human Tool Group (blue) and AI Tool Group (orange)
• Figure 5: Example of a structured user profile prompt provided to the LLM for in-context learning.