Harrison Chase of LangChain on Deep Agents, LangSmith, and Earning Trust | NVIDIA AI Podcast Ep. 297

2026-05-06 · Source: NVIDIA · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Advanced, extended

Summary

Harrison Chase, CEO and co-founder of LangChain, discusses the evolution of building applications with Large Language Models (LLMs), focusing on the concept of "Deep Agents." LangChain, founded three years ago, provides tools for developers to create complex, agentic systems around LLMs, with over a billion downloads. Deep Agents represent a new, general-purpose, model-agnostic, open-source agent harness that simplifies building autonomous LLM applications by providing common patterns for connecting to file systems, using sub-agents, and planning. This architecture, which powers systems like OpenClaw, allows LLMs greater autonomy in interacting with environments. Chase also highlights LangSmith for observability and evaluation in the agent development lifecycle (build, test, run, manage) and discusses the importance of secure runtimes like NVIDIA's OpenShell for managing agent skills and ensuring enterprise trust.

Key takeaway

For AI Architects designing and deploying LLM-powered solutions, you should prioritize adopting general-purpose agent harnesses like LangChain's Deep Agents to accelerate development and ensure scalability. Focus on integrating robust observability and evaluation platforms, such as LangSmith, from the outset to build trust and manage the iterative nature of agent development, especially when considering proactive, always-on enterprise agents. Re-evaluate existing agent architectures every nine months to capitalize on performance and scope improvements.

Key insights

Deep Agents provide a general-purpose, open-source harness for building autonomous, event-driven LLM applications with enhanced control and observability.

Principles

• Agent development is highly iterative.
• Observability and evaluation are crucial for agent trust.
• Coding models excel as general-purpose agents.

Method

LangChain's agent development lifecycle involves building with open-source frameworks (Deep Agents, LangGraph), then testing, running, and managing with LangSmith for observability and evaluation-driven development.

In practice

• Use LangGraph for controlled, directed workflows.
• Implement LangSmith for agent observability and testing.
• Consider open-source models for cost-sensitive, always-on agents.

Topics

• LangChain
• Deep Agents
• LangSmith
• Enterprise AI Adoption
• Agent Observability

Best for: AI Architect, AI Engineer, MLOps Engineer, Director of AI/ML

Related on AIssential

• Langfuse vs LangSmith: Two Competing AI Observability Platforms Compared — Effective AI agent debugging requires deep observability into internal decision processes, not just final outputs.
• Why AI Engineers Are Moving Beyond LangChain to Native Agent Architectures — LLM orchestration frameworks trade development speed for clarity, creating "abstraction debt" in production.
• Evaluating Deep Agents using LangSmith on AWS — Evaluating deep AI agents requires a multi-faceted approach combining diverse grading methods and evaluation patterns across the development lifecycle.
• How agent o11y differs from traditional o11y — Phil Hetzel, Braintrust — Agent observability requires specialized approaches due to LLM non-determinism, complex trace data, and diverse stakeholder involvement.
• How to Make a Coding Agent a General Purpose Agent - Harrison Chase — Coding agents, with robust harnesses and secure tool runtimes, form the foundation for scalable, general-purpose enterprise AI agents.
• Harness Engineering with LangChain DeepAgents and LangSmith — Harness engineering enhances LLM reliability by structuring the operational environment with prompts, tools, and middleware.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by NVIDIA.