Building your own Developer Advocate with Deep Agent!

Staying current with emerging frameworks is exhausting and can be swayed by bias. New repositories appear daily, each claiming to solve our problems better than the last. How do we separate the signal from the noise?

In this session, we'll walk through building a multi-agent research system using LangChain's DeepAgents framework. The system uses specialized SubAgents to evaluate technology viability: one tracks GitHub metrics (stars, commit velocity, contributor health), another analyzes issues and discussions for red flags, and a third synthesizes findings into actionable recommendations.

You'll learn how DeepAgents' built-in planning tools, filesystem backend, and sub-agent delegation handle the complexity of parallel research tasks. We'll cover practical patterns for context isolation between agents, when to spawn SubAgents vs. handle tasks inline, and how to persist research findings across sessions using composite backends.

Walk away with a working architecture you can adapt for your own technology vetting workflows.