What are Deep Agents? Explained fast
Read the Portuguese translation →
Author: LangChain · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: What are Deep Agents? Explained fast · Channel: LangChain · 01:06 · Watch on YouTube
Topics: AI agents, AI developer tools, AI safety
Summary
Section titled “Summary”Deep Agents is LangChain’s open-source harness for building complex agents, providing tools to manage context, persistence, actions, and boundaries. It includes file systems, skills, subagents, summarization, sandboxes, human-in-the-loop gates, and Agent Auth. The harness is customizable, allowing components such as the file system to be swapped without rebuilding the agent. LangChain Academy offers a free hands-on course about Deep Agents.
Key points
Section titled “Key points”- An agent is a model plus a harness.
- The harness controls what context the agent sees and when.
- Deep Agents includes tools for longer-running tasks, such as file systems, skills, subagents, and summarization.
- Sandboxes let agents run untrusted code safely.
- Human-in-the-loop gates sensitive actions before they go through.
- Agent Auth checks whether the agent is authorized to take an action.
- The file system, skills, and subagents are customizable or swappable.
- The same agent can run on a laptop, in a sandbox, or in the cloud without code changes.
- LangChain Academy offers a free Deep Agents course.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | What Deep Agents provides | The video introduces Deep Agents as an open-source harness and outlines its context-management tools and sandboxing. |
| 00:34 | Safety and customization | It explains human approval gates, Agent Auth, and how the harness components can be configured or swapped. |
| 01:06 | Course invitation | The video closes by pointing viewers to a free, hands-on Deep Agents course. |
Quotes
Section titled “Quotes”Deep Agents is LangChain’s open source harness for building complex agents. — 00:00
One goal of this harness is to control what context the agent sees and when. — 00:00
It also adds sandboxes so agents can run untrusted code safely. — 00:00
Then human-in-the-loop gates sensitive actions before they go through. — 00:34
Just configure the harness. — 00:34
Concepts and entities
Section titled “Concepts and entities”Deep Agents, LangChain, Harness, Agent Auth, LangChain Academy
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Deep Agents is LangChain’s open source harness for building complex agents. And an agent is just a model plus a harness. With the harness, the model can perceive, act, persist information, and enforce its boundaries. One goal of this harness is to control what context the agent sees and when. As an agent performs longer running tasks, it needs better context controlling tools like file systems, skills, subagents, and summarization. The Deep Agents harness ships with all of this out of the box. It also adds sandboxes so agents can run untrusted code safely.
00:34 Then human-in-the-loop gates sensitive actions before they go through. And finally, Agent Auth checks whether the agent is even authorized to take that action. Deep Agents ships with the batteries included, but it is fully customizable. The file system is swappable, so the same agent can run on your laptop, in a sandbox, or in the cloud without changing any code. Skills and subagents plug in the same way. No need to rebuild the entire thing. Just configure the harness. If you’re interested in learning more about the Deep Agents harness, LangChain Academy’s free Deep Agents course takes you through everything bit by bit.
01:06 The environment, backends, skills, subagents, memory, and so much more. Everything in this video, hands-on. See you there.