We’re releasing a new LangChain Academy course,  

LangChain Essentials, where you can  learn the basics of LangChain in an hour.

LangChain is the best way to get started  with generative AI. LangChain has had  

pre-built agents from very early in its  development. Together with the community,  

we’ve learned from production use-cases  and identified the essential components  

an agent needs. The `create_agent` abstraction  captures these qualities in their simplest form.

In this course, you’ll learn to build agents  with the create_agent abstraction. The pre-built  

agent is a ReAct-style agent, where the LLM  reasons and then acts using tools, in a loop.  

This architecture has proven to be both simple  and powerful. We wanted the agent to be easily  

scalable and resilient in the event of failures,  while also allowing for a human-in-the-loop in  

situations where you’d want human intervention.  To achieve this, we built on top of LangGraph.

A challenge in building any abstraction is  balancing flexibility with the benefits of being a  

pre-built abstraction. This agent can dynamically  select prompts and models, while optional  

middleware enables customization only when you  need it—using pre-built or custom implementations.

In the course, you’ll learn  to use many features of the  

create_agent abstraction as you build  an increasingly sophisticated agent.  

You’ll also use many of the building blocks  in LangChain, including messages, tools,  

and models. I hope that you can take the time  to learn more about LangChain in this course.