Title: AgentxGCore: Agentic AI for Next-Generation Mobile Core Network

Abstract

As emerging applications impose stricter demands and network operations grow increasingly intricate, the Sixth Generation (6G) of mobile networks—also known as Next Generation Mobile Networks (NextG)—is moving toward an AI-native Core Network (CN) architecture. The Third Generation Partnership Project (3GPP) has initiated this transition by introducing new functions to the cellular CN, marking the first step in integrating artificial intelligence (AI), analytics, and machine learning. However, these additions are currently limited by centralized management structures and significant operational complexity.

The advent of Large Language Models (LLMs) has ushered in a new phase for network management, one that leverages Intent-based Networking (IBN) to enhance orchestration capabilities. Within this context, Agentic AI and AI agents utilize Reasoning and Acting (ReAct) frameworks to continuously interact with network infrastructure using these high-level intents. While current state-of-the-art solutions primarily apply Agentic AI to reduce deployment and configuration hurdles within the CN, this study presents AgentxGCore. This framework introduces an Agentic AI-Native layer that extends the 3GPP architecture, facilitating a system that operates via existing APIs across the Beyond Next Generation Core (xGC) domain.

AgentxGCore establishes a closed-loop, AI-driven mechanism for continuous optimization, utilizing real-time data to achieve self-organization and self-adaptation. The system employs a multi-agent architecture comprising two distinct roles: a network planner agent, which visualizes the network status and formulates strategies to satisfy user intents, and a network executor agent, tasked with critiquing and implementing these plans. To verify the efficacy of this approach, the authors constructed a testing environment using an open-source CN, diverse datasets, and various LLMs.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC