What this does

LangGraph enables building stateful, cyclic agent workflows where nodes represent steps (LLM calls, tool execution) and edges define transitions. This guide builds a simple ReAct-style agent from scratch.

Steps

Define the agent state. Use TypedDict to specify the state schema.

from typing import TypedDict, Annotated, Sequence
from langgraph.graph import StateGraph, END
from langgraph.graph.message import add_messages
from langchain_core.messages import BaseMessage

class AgentState(TypedDict):
    messages: Annotated[Sequence[BaseMessage], add_messages]

Create the LLM node. Define a function that calls the LLM and returns the response.

from langchain_ollama import ChatOllama

llm = ChatOllama(model="llama3.2", temperature=0)

def call_model(state: AgentState) -> dict:
    response = llm.invoke(state["messages"])
    return {"messages": [response]}

Build the graph. Add nodes and edges.

workflow = StateGraph(AgentState)

workflow.add_node("agent", call_model)
workflow.set_entry_point("agent")

workflow.add_edge("agent", END)

app = workflow.compile()

Define a tool node (optional but common). For agents that call tools.

def tool_node(state: AgentState) -> dict:
    # Parse tool calls from last message, execute, return results
    return {"messages": [tool_result_message]}

workflow.add_node("tools", tool_node)
workflow.add_conditional_edges("agent", should_continue, {"continue": "tools", "end": END})

Invoke the agent.

from langchain_core.messages import HumanMessage

result = app.invoke({
    "messages": [HumanMessage(content="What is the capital of France?")]
})
print(result["messages"][-1].content)

Verification

python -c "
from langgraph.graph import StateGraph, END
from typing import TypedDict
class S(TypedDict):
    x: int
def inc(s):
    return {'x': s['x'] + 1}
g = StateGraph(S)
g.add_node('inc', inc)
g.set_entry_point('inc')
g.add_edge('inc', END)
app = g.compile()
r = app.invoke({'x': 0})
print(r['x'])
# Expected: 1
"

Common failures

Missing state key. The state dict must contain all keys defined in the TypedDict. Initialize with defaults in the first invocation.
No add_messages reducer. Without the reducer annotation, messages overwrite instead of appending.
Graph compilation fails without END. Every path must eventually reach END or a terminal node.
Version mismatch - The installed package or runtime differs from the command shown; check the version first and rerun the smallest verification command.
Local environment drift - Another service, virtual environment, model, or path is being used; print the active binary path and configuration before changing the guide steps.

Related guides

How to Define State Schema in LangGraph
How to Implement Tool Calling in LangGraph Agents