mkurman/langgraph

Overview

LangGraph builds stateful, multi-step agent workflows as graphs. Supports conditional routing, human-in-the-loop checkpoints, persistent state, streaming, and multi-agent orchestration. The graph-based design enables complex, controllable agent behavior.

Installation

bash
uv pip install langgraph

Simple Graph

python
from typing import TypedDict
from langgraph.graph import StateGraph, END

class AgentState(TypedDict):
    messages: list
    next_step: str

def research(state):
    return {"messages": state["messages"], "next_step": "write"}

def write(state):
    return {"messages": state["messages"], "next_step": "review"}

def review(state):
    return {"messages": state["messages"], "next_step": "__end__"}

graph = StateGraph(AgentState)
graph.add_node("research", research)
graph.add_node("write", write)
graph.add_node("review", review)
graph.set_entry_point("research")
graph.add_edge("research", "write")
graph.add_conditional_edges("write", lambda s: s["next_step"])
graph.add_edge("review", END)
app = graph.compile()

result = app.invoke({"messages": [], "next_step": ""})

References

• LangGraph docs
• LangGraph GitHub