A collection of concepts, examples, and projects to learn and apply LangGraph — from fundamentals to building stateful, multi-step agentic workflows.
LangGraph is a framework built on top of LangChain that allows you to create graph-based workflows for LLMs.
Instead of simple sequential chains, LangGraph lets you define stateful, branching, and looping logic for agents.
This makes it possible to build applications such as:
- Conversational agents with memory
- Multi-agent systems that collaborate
- Complex decision-making workflows
- RAG pipelines with dynamic control
While LangChain provides building blocks (chains, agents, tools), it can be hard to manage complex, multi-step workflows.
LangGraph solves this problem by:
✅ Defining workflows as graphs of states & edges
✅ Supporting branching, loops, and conditional flows
✅ Adding persistence & state management
✅ Making agents more reliable and controllable
With LangGraph, you can design applications where the LLM is not just reacting, but moving through a defined graph of logic.
-
State
- Stores variables across the workflow (e.g., user inputs, retrieved docs).
- Example: Conversation history, query context.
-
Nodes
- Units of execution (can be functions, chains, or agents).
- Example:
collect_info → call_tool → summarize.
-
Edges
- Define transitions between nodes (sequential, conditional, or looping).
- Example:
if success → next stepelseretry.
-
Graph Execution
- Workflows are run like finite state machines, ensuring predictable flow.
-
Persistence
- Ability to pause, resume, or replay workflows with stored state.
# Clone the repo
git clone https://github.com/alihassan-coder/langgraph-concepts-projects.git
cd langgraph-concepts-projects