Skip to content
Conference Open access

Autonomous Multi-Agent Systems Orchestrated via n8n: Infrastructure Bottlenecks and Self-Healing Architectures

Aug 2026 · 2026 International Conference on Secure Information Systems and Technologies (ICSIST) · pp. 681-689 · 0 citations · 16 references

TL;DR

A Split-Stack Orchestration Architecture is proposed that allocates strategic reasoning, execution, and low-cost repair tasks to separate model tiers, reducing sanitization cost by an estimated 99.2% while decoupling execution-tier rate-limit exhaustion from strategic-tier availability.

Abstract

The transition from single-shot generative models to autonomous, goal-directed agents represents a structural departure from fixed-pipeline automation. Low-code orchestration platforms such as n8n increasingly supply the operational substrate for this transition, handling function invocation, state persistence, and inter-agent communication while the reasoning itself remains delegated to a large language model (LLM). This paper investigates, through two controlled case-study experiments built on a Gemini 2.0 Flash agent orchestrated in n8n, whether the sustainable operating depth of such systems is bounded by the reasoning capacity of the model or by the surrounding infrastructure. We introduce the Agentic Depth Coefficient (ADC), a formal metric for the number of consecutive successful reasoning-action cycles an agent completes before terminal failure, and the Agentic Thirst equation, which expresses the cumulative API demand of a multi-turn agentic task as a function of ADC, tool calls, and reasoning sub-steps. In Experiment 1, an openended literature-synthesis task terminated at ADC = 6 due to an HTTP 429 rate-limit response rather than any indication of reasoning failure, evidencing that free-tier infrastructure quotas, not model cognition, defined the operating ceiling. In Experiment 2, a secondary Healer agent achieved full recovery from all deterministic parsing failures induced by semantically valid but syntactically irregular natural-language input, a behavior we formalize as Semantic Override. Based on these findings we propose a Split-Stack Orchestration Architecture that allocates strategic reasoning, execution, and low-cost repair tasks to separate model tiers, reducing sanitization cost by an estimated 99.2% while decoupling execution-tier rate-limit exhaustion from strategic-tier availability. As single-instance case studies, these experiments are intended to demonstrate the existence and mechanism of infrastructure-bound and semantically self-healing behavior rather than to establish statistically generalizable failure rates; we discuss this scope explicitly in Section IX and outline the replication needed to establish the latter.

Read PDF

Similar papers

#artificial intelligence Preprint Oct 2026

MAS-Gate: Unlocking Multi-Agent System Lifecycle via Node-Centric Quantification

Large language model (LLM)-based multi-agent systems (MAS) can solve complex problems beyond the reach of a single agent by decomposing tasks and coordinating complementary capabilities. Despite extensive research on MAS, existing work largely focuses on specific aspects or individual phases of their development and op...

He-Hai Lin, Zi-Xuan Wang, Ming Yan et al. · 1 citation
Preprint Sep 2026

Uncertainty-Aware Resource Orchestration for Agentic AI Systems

Agentic artificial intelligence (AI) systems powered by large language models (LLMs) execute as autonomous multi-step pipelines that dynamically invoke tools, integrate observations into memory, and iterate until a task is resolved or a resource budget is exhausted. Unlike classical workflows with fixed graphs, the agg...

Masoud Shokrnezhad, T. Taleb · 0 citations
Preprint Oct 2026

DynaFlow: Learning Free-Form Multi-Agent Orchestration

Multi-agent systems often rely on predefined collaboration structures, limiting their ability to adapt to diverse task requirements. We introduce DynaFlow, a framework that integrates free-form executable workflow synthesis into the action space of the main agent. The agent decides when collaboration is needed and gene...

Ru-Hao Liu, Qi Li, Shi-Zun Wang et al. · 1 citation
Preprint Oct 2026

AgentSoV: Virtual Sensing for LLM Multi-Agent Systems via Stream of Variation

AgentSoV reframes agentic execution as a dynamic system to model rather than merely text to inspect, providing insights for MAS structure and component design and demonstrating a unified approach to fault localization, information influence, and observability.

Cong-Jing Zhang, Si-Yuan Xu, Yu Bai et al. · 0 citations
#artificial intelligence Preprint Oct 2026

MAS-Loop: Beyond One-Shot Orchestration towards Self-Evolving Multi-Agent Systems

Training-time multi-agent system (MAS) design seeks to learn reusable orchestration policies, yet prevailing approaches remain largely one-shot: an Orchestrator generates a system but only observes and learns from its terminal outcome. This paradigm discards execution trajectories that expose system-level failures and...

He-Hai Lin, Qi Yang, Cheng-Wei Qin · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.