What is Learn LangChain?
Learn LangChain is a comparative studio for the LangChain stack. Same Tuesday ticket, four runtimes: a chain, create_agent, create_deep_agent, and a LangGraph.
You pick a job and a runtime and run it. A chain searches once. A loop can search again. A harness brings a filesystem you may not need. A graph can wait for a person. The goal is to get someone from "I have heard of LangChain" to "I can pick a shape for this ticket."
Why We Built It
LangChain is a pile of product names. Chain, LangChain, Deep Agents, LangGraph, LangSmith, Agent Server, Fleet, Engine, LLM Gateway. The docs will give you all of them. The thing that actually matters is the job: which runtime for this Tuesday ticket.
A PDF one-pager is a chain. Docs that have to look again are a loop. A rental that waits for a person is a graph. A QBR that needs files and subagents is a harness. A morning digest that a human should see before it sends is a harness plus a pause.
We built Learn LangChain so people can run those jobs instead of only reading the catalog.
A Hands-On Way to Compare Four Runtimes
This is not a slide deck about the stack.
You pick a job, pick a runtime, press Play. The matrix is five tickets by four machines. After the run, a Smith drawer shows a simulated trace, eval, deploy, and gateway. No API key. Nothing calls a model. Sign in with your En Dash account if you want progress and places on another machine.
Built Around Jobs, Then the Stack
Two doors.
Jobs are the tickets: PDF one-pager, grounded docs, rental gate, QBR, morning digest. Each one has a runtime that fits and one that does not.
Studio is the matrix. Same ticket as a chain, a loop, a harness, and a graph. The glossary on the landing is what the words mean here, not a product brochure.
Useful for Engineers Choosing a Shape
This is for people who already ship LLM features and are staring at four product pages.
If the steps are known and the work never comes back, stay on a chain. If the model needs to try again, use create_agent. If the job is long and messy enough to need files and subagents, use Deep Agents. If you need a custom mix of code, one LLM, and a human pause, draw a graph. Learn LangChain is the place to see those four next to each other.
Built to Make the Idea Click
Some product ideas fit in one sentence. This stack needs the same job, four times.
Learn LangChain exists so more people can see which runtime fits, which one is overkill, and which one cannot do the job at all.
Not official LangChain. A community resource from En Dash, same shop as Learn LangGraph, Learn AgentCore, html-in-canvas.dev, and learn-pretext.com.