2 Justification

Agentic AI is becoming increasingly important as industries move from passive analytics to systems that can act autonomously and adapt in real time. AI agents reduce manual effort, speed up decisions, and optimize processes in complex environments like networks, logistics, and automation.

Unlike traditional AI that only provides insights, AI agents make decisions and take actions to achieve specific goals. Powered by technologies like large language models (LLMs), they can understand context, plan next steps, and interact with other agents and their environment enabling a new level of intelligence and automation in digital ecosystems.

To unlock this potential, AI agents must interact not only with other agents but also with non-AI systems including information services (e.g., search and retrieval), physical-world interfaces (e.g., sensors and actuators), network infrastructure, and business applications and workflows.
This is where the new Model Context Protocol (MCP) plays a critical role: it provides a structured and standardized way for agents to exchange context and commands with existing digital platforms, enabling seamless integration across AI and non-AI domains.

Model Context Protocol (MCP) is rapidly gaining traction across the industry, with adoption by leading agent platforms and integration efforts from major technology players.

In this work item, we propose to explore the interworking of MCP with oneM2M enabling the broad ecosystem of IoT devices supported by this standard to be easily extended with agent-based intelligence and capabilities, while also allowing oneM2M to enhance the capabilities of agents.