Webinar: Agent Observability Powers Agent Evaluation

​Join Harrison Chase, Co-founder and CEO at LangChain, for a technical deep dive into why building reliable AI agents requires new approaches to observability and evaluation.

​AI agents do not fail like traditional software. When an agent takes hundreds of steps, repeatedly calls tools, updates state, and still produces the wrong result, there is no stack trace to inspect. Nothing crashed. What failed was the agent’s reasoning.

​In this webinar, we’ll explore why observability and evaluation for agents are fundamentally different from what most software teams are used to, and why traces become the primary source of truth when you start building agentic systems.

​We’ll cover how debugging shifts from finding a broken line of code to understanding where an agent’s reasoning went off track, why evaluating agents requires looking at trajectories and decisions rather than just final outputs, and how production traces enable both offline and online evaluation.

​We’ll also walk through the core primitives of agent observability, including runs, traces, and threads, and show how teams use them together to iteratively improve agent behavior with LangSmith.

​This session is for engineers and teams building agents who are encountering non-deterministic behavior and want a clearer mental model for how to debug, evaluate, and ship agents that actually work.

​Agenda:

​12:00pm: Presentation by Harrison

​12:30pm: Q&A session

​About LangChain:

​LangChain provides the agent engineering platform and open source frameworks to help companies ship reliable agents. LangSmith powers top engineering teams like Cisco, Klarna, Clay, Blackrock, LinkedIn, and more. Observe, evaluate, and deploy agents with LangSmith, our comprehensive platform for agent engineering. Learn more: https://www.langchain.com/