Fraunhofer FOKUS investigates agentic AI for 5G/6G core networks, where AI agents become active components of network operation. Open6GCore provides an executable environment for integrating AI agents with network data, network functions and control mechanisms and for experimentally evaluating autonomous and collaborative network operation.
Fraunhofer FOKUS has been demonstrating the Open6GCore Agentic Core publicly since MWC Barcelona in March 2026, reflecting an early move from agentic-network concepts toward running implementations and end-to-end experimentation.
The Agentic Core architecture provides AI agents with access to the Open6GCore shared data backbone and exposes network information and capabilities through Model Context Protocol (MCP) interfaces. A2A communication enables agents to discover and interact with other agents, providing a basis for distributed decision making and multi-agent coordination.
Our research investigates different ways of integrating agents into mobile networks: agents can assist existing network functions, interact with several functions through common interfaces, operate as network functions themselves, or form
a more deeply integrated agentic core architecture. This enables research on autonomous decision making, coordinated network adaptation, and the interaction between AI agents and conventional network control.
Research Topics
- Agentic AI for 5G/6G core networks
- AI agents interacting with network functions
- MCP-based access to network data and capabilities
- A2A communication and multi-agent coordination
- Autonomous network decision making
- Agents as or within network functions
- Distributed agent architectures
- Closed-loop network adaptation
- Security, trust and control of autonomous agents
Open6GCore provides the network data, interfaces, control capabilities and running 5G/6G environment required to move agentic networking from architectural concepts to working implementations, demonstrations and end-to-end experiments.
