Industry intelligence and market context for members.
Trendline: MCP and LSO — How AI Agents Are Learning to Speak Network
A new interaction model is taking shape across software, enterprise platforms, and network operations. Model Context Protocol (MCP), originally developed to give AI structured access to tools and data, is quickly emerging as a de facto integration standard.
For the NaaS ecosystem, MCP is not simply another API framework. It introduces a model where AI agents can discover, invoke, and coordinate network capabilities with context and intent, rather than through rigid, point-to-point integrations.
Why MCP Matters for Network Automation
Traditional network automation relies on well-defined APIs. It’s powerful, but dependent on precise, human-defined logic to trigger and sequence operations.
MCP shifts this model by enabling agent-based interaction with APIs: discovering available capabilities, selecting appropriate actions, and executing workflows based on intent rather than explicit instruction.
For NaaS providers, the interface is no longer just human-to-system; it is increasingly agent-to-network. Provisioning, performance monitoring, fault diagnosis, and service modification become candidates for AI-driven orchestration when network APIs are exposed in an MCP-compatible format.
“By combining standardized LSO APIs with MCP support, Mplify enables AI agents and large language models to interact more directly with networking infrastructure, helping providers reduce service complexity, accelerate decision-making, and move closer to autonomous operations.”
— Pascal Menezes, CTO, Mplify
Kylie in Practice
The Kylie SDK release translates this model into operational capability. With MCP support across the LSO API portfolio—Sonata, Cantata, Allegro, Interlude, and Legato—network functions exposed through standardized interfaces become directly accessible to AI agents.
The LSO API Blending Tool extends this by enabling providers to combine core LSO APIs with product-specific schemas, including IP, Carrier Ethernet, wavelengths, data center cross-connects, CAMARA Quality on Demand, and SD-WAN, and generate MCP-compatible interfaces.
For operations teams, this reduces integration complexity and accelerates deployment. For the broader ecosystem, it establishes a scalable approach for how standardized APIs and AI-native interaction frameworks can operate together across multi-provider environments.
From Managed Resource to Active Participant
As agentic AI systems become persistent and distributed—operating across cloud, edge, and network domains, the role of connectivity is changing.
Networks are moving from managed resources to active participants in the system.
LSO APIs, combined with standardized data models, provide the foundation. MCP defines the interaction model. Together, they enable intelligent systems to access, coordinate, and operate network infrastructure at scale.
Additional Resources
