As networks become more distributed, dynamic, and software-driven, the industry is reaching a clear turning point: automation alone is no longer enough. The next phase is agentic AI: systems that can reason, act, and collaborate across complex environments. That is the focus of our recent white paper, Architecting Autonomy: The Convergence of Agentic AI and Open Source Networking, authored by Sana Tariq of LFN member organization TELUS, Ranny Haiby (CTO of Networking, Edge and Access) and Arpit Josphipura (GM of Networking, Edge, IoT) of the Linux Foundation.
The paper makes the case that agentic AI is not just another layer of tooling. It represents a broader architectural shift for networking, one that moves the industry beyond static automation toward more adaptive, intent-driven, and ultimately more autonomous operations. At the same time, it argues that this evolution must remain grounded in open standards, interoperability, and vendor-neutral collaboration.
At the heart of the paper is a simple but powerful idea: the future of networking depends on advancing both “AI for Networks” and “Networks for AI”:
- On one side, AI agents can help optimize, automate, troubleshoot, and remediate networks.
- On the other, networks themselves must evolve to support the demanding requirements of AI workloads, including distributed training, inference, edge discovery, and SLA-aware connectivity.
The paper positions agentic AI as the control plane that helps connect these two transformations.
A major theme is the importance of a shared framework for how AI agents interact with tools, data, and infrastructure. It points to the Linux Foundation’s Agentic AI Foundation and the growing role of agent communication standardization efforts like the Model Context Protocol (MCP) as a vendor-neutral foundation for agent orchestration. In this model, LF Networking projects can expose capabilities through standardized, machine-consumable interfaces, allowing AI agents to work across environments in a more predictable, secure, and scalable way.
The paper also highlights why open source networking (and adjacent) projects are especially well positioned for this shift. Projects such as Nephio, ONAP, CAMARA, Sylva, and SONiC already provide critical pieces of the networking and cloud stack. The opportunity now is to make those capabilities more discoverable and more usable by AI agents, whether for intent-based orchestration, closed-loop assurance, developer access to network APIs, or infrastructure optimization for AI-native services. The Essdum project plays a key role in this transformation.
The paper highlights practical use cases: how agentic AI can enable closed-loop assurance, natural-language-to-network intent mapping, multi-domain service provisioning, operational troubleshooting, and more dynamic exposure of network capabilities through APIs. In other words, this is not a theoretical exercise. It is a roadmap for how open networking communities can translate AI’s promise into real operational value.
Another key highlight is that autonomy is not the same as unbounded automation. Instead, The whitepaper clearly defines principles for applying AI to networking such as deterministic execution, governed autonomy, separation of reasoning and execution planes, first-class observability, secure and auditable access, discoverable capabilities, backward compatibility, and open governance. These principles form LFN’s proposed Agentic AI Readiness Charter, giving projects a practical foundation for adopting agentic capabilities responsibly.
As agentic systems become a larger part of digital infrastructure, networking cannot afford to evolve in silos. Open source communities have a critical role to play in shaping how autonomy is built into networks; not as proprietary black boxes, but as interoperable, auditable, and extensible systems developed in the open. The paper offers an important framework for that work and a timely call for the ecosystem to engage.
Read the full white paper, Architecting Autonomy: The Convergence of Agentic AI and Open Source Networking, to explore the architecture, use cases, and readiness principles shaping the next chapter of AI-native networking.
