Agentic-NWDAF: Enabling intent-driven agentic intelligence for autonomous 6G network analytics

The transition from 5G to 6G envisions Operations Support Systems (OSS) that embody Zero-touch Service Management (ZSM) and Intent-Based Networking (IBN) as core enablers of autonomous network operation. Ongoing standardization efforts by the 3rd Generation Partnership Project (3GPP) and the European Telecommunications Standards Institute (ETSI) have defined how OSS interfaces with Network Data Analytics Functions (NWDAF) through structured, service-based APIs that support performance assurance, fault detection, and closed-loop automation. However, current approaches remain largely procedural, limited to static analytics subscriptions and rule-based triggers, and thus fall short of enabling intent-aware, adaptive, and democratized OSS–NWDAF interactions. To overcome these limitations, this paper introduces Agentic-NWDAF, an intent-aware orchestration layer that translates natural language OSS intents into executable NWDAF service calls. Agentic-NWDAF leverages Agentic Artificial Intelligence (AI) and the Model Context Protocol (MCP) to enable natural language understanding, adaptive analytics orchestration, and autonomous decision-making through collaborative AI agents. We validated AgenticNWDAF through an extensive benchmark of 100 NWDAF intents derived from 3GPP specifications, demonstrating high accuracy, strong generalization, and reliable performance under LLM-as-a-Judge evaluation.