Monitoring Agentic Applications

Learn about observability and monitoring challenges for agentic AI applications in this conference talk from the Linux Foundation. Discover how the evolving landscape of generative AI technologies creates complex management challenges, particularly when applications include multiple agents and tools that may be embedded or remote over MCP (Model Context Protocol). Explore how application quality, reliability, and performance depend on dynamic decisions for agent delegation and tool selection. Get introduced to Project Monocle, a community-driven open source project under Linux Foundation AI&Data that enables developers to trace their applications and agents seamlessly with minimal code changes. Understand how Monocle generates traces that outline agent and tool operations while providing a consistent format to describe entities like agents, tools, and LLMs, as well as events such as prompts, responses, and agent delegation. See how this solution can be integrated in both personal development environments and cloud deployments, addressing the critical gap between development prototypes and production-ready agentic applications.