Agentic AI Architecture and Multi-Agent Orchestration: From Static Language Models to Autonomous Execution Ecosystems

Paradigm Shifts, Reasoning Loops, Tool Augmentation, and Safety Guardrails

Authors

  • Gusti Tri Prayoga, S.Kom. Department of Informatics Engineering, Universitas Pahlawan Tuanku Tambusai, Pekanbaru, Indonesia
  • Hermes Engineering Division Tunas Melayu Digital, Artificial Intelligence & Autonomous Systems Research Group

Abstract

The evolution of Large Language Models (LLMs) has transitioned from passive prompt-and-response paradigms toward proactive, goal-driven autonomous systems known as Agentic AI. This paper presents a rigorous investigation into the structural architecture, coordination mechanics, and runtime governance of tool-augmented cognitive agents. We identify four fundamental pillars: dynamic task decomposition, external tool execution with deterministic grounding, hierarchical episodic-procedural memory retention, and distributed multi-agent orchestration. We systematically analyze recursive hallucination cascades, context window saturation, and sandboxed execution isolation. Finally, we showcase an end-to-end multi-agent deployment coordinating real-world software engineering pipelines and digital infrastructure operations.

Author Biography

  • Gusti Tri Prayoga, S.Kom., Department of Informatics Engineering, Universitas Pahlawan Tuanku Tambusai, Pekanbaru, Indonesia

    Alumnus of Informatics Engineering (2022) at Universitas Pahlawan Tuanku Tambusai; Full-Stack Software Developer & AI Systems Practitioner.

References

Downloads

Published

2026-11-05

Issue

Section

Original Research Articles (General Multidisciplinary)

License

Creative Commons Attribution 4.0 International (CC BY 4.0)
Open Access Article
This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.
Screening Plagiarism PASSED
Similarity Index: < 15% (Turnitin Verified). Peer-reviewed with Zero Plagiarism Policy.
Official Letter of Acceptance (LoA) PDF

How to Cite

Citation Style
Loading citation...
Download .RIS

How to Cite

Agentic AI Architecture and Multi-Agent Orchestration: From Static Language Models to Autonomous Execution Ecosystems: Paradigm Shifts, Reasoning Loops, Tool Augmentation, and Safety Guardrails. (2026). Journal of Tunas Melayu Digital, 1(1), 1-6. https://jutuma.org/index.php/tmd/article/view/agentic-ai-architecture
JTMD • Vol 1 No 1 Agentic AI Architecture and Multi-Agent...
Share: