From Protocol to Practice: Nodes, Not Agents in Human-in-the-Loop LLM Swarms — The MSCFT Ecosystem and its No-Code Templates (MSCFT, SENTINEL, FORGE, MATHEMATICS) for Forecasting, Coding, Security, and Mathematics

This paper, *From Protocol to Practice: Nodes, Not Agents in Human-in-the-Loop LLM Swarms — The MSCFT Ecosystem and its No-Code Templates for Forecasting, Coding, Security, and Mathematics*, is the second in the MSCFT series. It builds on the first paper, *From Coordination Failure to Scalable AI Swarms: The MSCFT Protocol for Structured Forecasting and Multi-Agent Alignment among Large Language Model Systems* (https://osf.io/gbhu3).

The work presents four no-code templates — MSCFT, SENTINEL, FORGE, and MATHEMATICS — implemented as custom GPTs to demonstrate a secure, scalable, and consensus-driven node-based swarm architecture. Each template defines a specific domain: MSCFT for forecasting, SENTINEL for security review, FORGE for code generation, and MATHEMATICS for quantitative reasoning.

By operationalizing the MSCFT framework, this research shows how large language models can be transformed from unstructured predictors into disciplined, human-in-the-loop instruments. It emphasizes nodes (bounded, role-specific participants) instead of “agents,” reinforcing structure, reproducibility, and oversight in multi-model systems.