In this session, we dive into the real-world economics of running AI agents in production.

Starting with a simple agent built using Google’s Agent Development Toolkit and Google Programmable Search for LLM web access, we walk through how to instrument it with open-source tool Weave for full tracing and insight into every single LLM call.

Learn how different components - LLMs, tool calls, and knowledge libraries - impact cost and performance.

The audience walks away with the tools to measure, debug, and optimize their agent workflows at scale.