Abstract for: Advanced Agentic System Dynamics: Orchestrating Causal Reasoning with MCP and Frontier Models

In 2026, as organizations deploy autonomous agentic workforces to manage complex systems, the primary risk has shifted from text generation to "causal blindness." This advanced workshop demonstrates how System Dynamics serves as the essential "reasoning guardrail" for AI agents, providing the mathematical structure necessary to prevent hallucinations in strategic decision-making. Participants will explore a bleeding-edge technical stack centered on the Model Context Protocol (MCP), which enables frontier LLMs to directly query and manipulate simulation models via a standardized JSON-RPC interface. We will deep-dive into the Validatable Model Standard (VMS), a Python-based framework that ensures agent-generated code remains structurally transparent and auditable. Through a high-speed, interactive "Agentic War Room," we will implement multi-agent workflows—such as collaborating Purchasing and Finance agents—that use "Simulation-Driven Development" to validate actions before deployment. Attendees will live-configure connections between reasoning models and VMS engines to experience real-time structural recovery and discuss "Agentic Governance" based on simulation confidence scores. This session is designed for practitioners at the absolute intersection of system dynamics, software engineering, and AI research.