Complete Deep Agents Course With Langchain In 3 Hours

Why It Matters

Deep agents give businesses the ability to automate intricate, multi‑stage tasks with coordinated AI sub‑agents and persistent memory, dramatically expanding the scope of practical AI deployments.

Key Takeaways

• Deep agents add planning, sub‑agents, and persistent memory.
• Shallow agents rely on single LLM‑tool loop without decomposition.
• React agents iterate tool calls but still lack structured planning.
• Core deep‑agent components: planner, sub‑agents, system prompt, file system.
• Example: Cloud‑Code uses to‑do list, sub‑agents, and shared storage.

Summary

The video introduces "deep agents," a next‑generation AI architecture that moves beyond the simple LLM‑tool loops of traditional agents. Krishna contrasts shallow agents—single‑step LLM calls to external tools—with more advanced React agents that can iterate tool calls but still lack explicit planning and memory.

Key insights include the limitations of shallow agents: no planning, inability to decompose complex queries, and minimal context retention. React agents improve by allowing multiple tool interactions, yet they remain shallow because they lack structured planning, state management, and persistent memory. Deep agents address these gaps with four core components: a planning module, sub‑agents that execute task fragments, a system prompt that defines behavior, and a shared file system for persistent state.

Krishna demonstrates the concept using Cloud‑Code as a real‑world deep research agent. The system prompt defines the agent’s role, the planner creates a to‑do list (e.g., booking a Paris trip), sub‑agents handle each step, and a shared file system stores intermediate results. A blog‑research example further shows how sub‑agents can parallelize internet searches, paper retrieval, writing, and copyright checks.

The implication is clear: deep agents enable enterprises to automate complex, multi‑step workflows with coordinated sub‑agents and memory, unlocking higher productivity and more sophisticated AI‑driven services.

Original Description

github: https://github.com/krishnaik06/Deep-agents-With-Langchain

The easiest way to start building agents and applications powered by LLMs—with built-in capabilities for task planning, file systems for context management, subagent-spawning, and long-term memory. You can use deep agents for any task, including complex, multi-step tasks.

Deep Agents is an “agent harness”. It is the same core tool calling loop as other agent frameworks, but with built-in capabilities that make agents reliable for real tasks.

Timestamp

00:00:00 Introduction

00:02:31 what are Deep Agents

00:16:20 Project Implementation

00:21:56 Building Deep Agents With Langchain

00:43:45 Customize Deep Agents

00:46:04 Deep Agents vs Claude Agent SDK

00:59:34 Deep Agents Backend Agents.md

01:19:05 Context Engineering And Memory In Deep Agents

01:53:15 Skills In Deep Agents

02:17:55 Subagents In Deep agents

02:34:22 End To End Deep Agents Project

----------------------------------------------------------------------------------------------------

Learn from the best to be the best

https://krishnaik.in/liveclasses