AgentEdge: Agentic AI for Service Orchestration in the Edge-Cloud Continuum

The recent emergence of Generative Artificial Intelligence (GenAI), particularly Large Language Models (LLMs), has enabled the development of autonomous agents capable of making complex decisions and coordinating actions across distributed systems. This evolution toward agentic AI systems presents significant opportunities for sixth generation (6G) mobile networks, which require orchestration systems that can leverage distributed intelligence across complex edge-cloud environments. This article introduces AgentEdge, the first multi-agent system specifically designed for edge-cloud service orchestration that distributes intelligence across specialized autonomous agents, while ensuring system-wide coordination. AgentEdge overcomes the limitations of traditional trial-and-error agentic AI methods by introducing the ActSimCrit (Action-Simulation-Critic) mechanism, i.e., instead of relying on real-world experimentation, ActSimCrit simulates and critically evaluates each orchestration decision before execution. This ensures reliability across the stack-from infrastructure adjustments to high-level application policies such as admission control-making it suitable for critical service environments. In addition, we list real-world use cases where AgentEdge can improve stability and performance, while outlining key practical challenges and open research directions for deploying the framework in large-scale, production-grade network environments. Experimental validation demonstrates AgentEdge's substantial benefits, achieving a 3.6× improvement in orchestration decision success rates compared to single-agent approaches while reducing API overhead by 2.8×.