AI AGENTS
Deep Agents
by LangChain
Overview
Open-source, production-oriented agent harness with planning, subagents, filesystem tools, persistent memory, and human approval controls.
Details
Deep Agents is LangChain’s open-source, production-oriented agent harness for building agents that can plan work, delegate to subagents, use filesystem tools, retain persistent memory, and incorporate human approval controls. Official Python documentation and API reference cover agent, middleware, filesystem, memory, skills, and subagent modules. LangChain also documents UI patterns for real-time subagent streams, task progress, and sandbox experiences.
When to Use
Use it when an agent needs to plan multi-step work and delegate portions of that work to subagents. Use it when workflows need filesystem access persistent memory or human approval controls. Use it when building a tool-calling agent and monitoring subagent activity or task progress in a UI.
Getting Started
- Install the Python `deepagents` package using the official quickstart.
- Create a tool-calling Deep Agent following the quickstart.
- Run the documented research workflow
- then review the overview and API reference for filesystem
- memory
- skills
- middleware
- and subagent modules.
Key Features
- •Planning and subagent support.
- •Filesystem tools and persistent memory.
- •Human approval controls.
- •Python package with agent
- •middleware
- •filesystem
- •memory
- •skills
- •and subagent modules.
- •UI patterns for real-time subagent streams
- •task progress
- •and sandbox experiences.
Capabilities
- •agent orchestration
- •planning
- •subagents
- •tool calling
- •filesystem access
- •persistent memory
- •human approval workflows
Last updated Sep 10, 2026