MCP-enabled AI automates IPoDWDM network lifecycle end-to-end
Live demo shows vendor-agnostic closed-loop control using GNPy and telemetry.
A team of researchers led by Chunmin Xia has introduced an MCP-enabled agentic AI architecture for fully autonomous control of IPoDWDM optical networks. The system, accepted for demo at the European Conference on Optical Communication (ECOC 2026), enables vendor-agnostic multi-layer lifecycle automation. By leveraging an agentic AI framework with MCP (likely a control protocol), the architecture can manage the entire network lifecycle—from provisioning to optimization—without human intervention.
The demo validates the approach on a real testbed, using GNPy for physical layer modeling and real-time telemetry for closed-loop feedback. This allows the AI agents to automatically detect performance degradation, reroute traffic, and optimize network parameters. The vendor-agnostic design means it works across different hardware vendors, a key advantage for operators managing multi-vendor infrastructures. The results represent a significant step toward self-healing, self-optimizing optical networks, reducing operational overhead and improving reliability.
- MCP-enabled agentic AI provides end-to-end lifecycle automation for IPoDWDM networks.
- System is vendor-agnostic and uses GNPy plus telemetry for multi-layer closed-loop control.
- Validated on a real testbed and accepted for demo at ECOC 2026.
Why It Matters
Network operators can achieve self-healing, self-optimizing optical networks with reduced manual intervention.