Building & Observing a Deep Agent for Email Triage with LangSmith

In this video, we walk through how to build and observe a deep agent using LangSmith.

We’ll build a simple email assistant that reads incoming emails and decides how to handle them — triage, respond, or take action — using a prompt-driven approach.

You’ll learn:

How to define a deep agent in a single file

• Why most agent complexity lives in the system prompt (not the architecture)
• How to encode rules, context, and decision criteria into prompts
• How to use LangSmith to observe, validate, and debug agent behavior

This walkthrough is useful if you’re building longer-running agents and want confidence that your agent is doing the right thing at each step.

- Learn more about LangSmith: https://docs.langchain.com/langsmith
- Learn more about debugging deep agents: https://blog.langchain.com./debugging-deep-agents-with-langsmith/
- Learn more about agent engineering: https://blog.langchain.com/agent-engineering-a-new-discipline/