SAGE: Sequential Agent Goal Execution Protocol for Multi-LLM Workflow Management

Muhammad Saim

Keywords: Multi-LLM Workflow Orchestration, Workflow Automation, Prompt Decomposition, Model Routing

TL;DR: A modular protocol designed to orchestrate complex, multi-step AI workflows by decomposing user prompts into validated, goal-driven sub-tasks.

Abstract: SAGE (Sequential Agent Goal Execution Protocol) is a modular protocol designed to orchestrate complex, multi-step AI workflows by decomposing user prompts into validated, goal-driven sub-tasks. Each sub-task is dynamically routed to the most suitable language model, executed, evaluated, and aggregated into a coherent final response. While the current demonstration leverages local language models via Ollama, SAGE is architected as an extensible and flexible framework, abstracting the workflow logic from any specific model provider or deployment environment

Submission Number: 65