How to beginner · 3 min read

How to build a simple LangGraph agent

Q: How to build a simple LangGraph agent

Use the langgraph package to define a StateGraph with typed states and nodes as functions. Compile the graph and invoke it with initial state data to run a simple LangGraph agent.

Quick answer

Use the langgraph package to define a StateGraph with typed states and nodes as functions. Compile the graph and invoke it with initial state data to run a simple LangGraph agent.

PREREQUISITES

Python 3.8+
pip install langgraph
Basic Python typing knowledge

Setup

Install the langgraph package via pip and prepare your environment.

bash

pip install langgraph

output

Collecting langgraph
  Downloading langgraph-0.1.0-py3-none-any.whl (10 kB)
Installing collected packages: langgraph
Successfully installed langgraph-0.1.0

Step by step

Create a typed State dictionary, define a node function that updates state, build a StateGraph, add nodes and edges, compile the graph, and invoke it with initial state.

python

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

# Define the state schema
class State(TypedDict):
    messages: list[str]

# Define a node function that appends a response
def my_node(state: State) -> State:
    return {"messages": state["messages"] + ["Hello from LangGraph agent!"]}

# Create the graph and add nodes
graph = StateGraph(State)
graph.add_node("my_node", my_node)
graph.set_entry_point("my_node")
graph.add_edge("my_node", END)

# Compile the graph into an app
app = graph.compile()

# Invoke the app with initial state
result = app.invoke({"messages": ["Start"]})
print(result["messages"])

output

['Start', 'Hello from LangGraph agent!']

Common variations

Use multiple nodes and edges to create complex flows.
Persist state with a checkpointer for long-running agents.
Integrate with LLMs by calling AI APIs inside node functions.

python

from langgraph.graph import StateGraph, END
from langgraph.savers.memory import MemorySaver
from typing import TypedDict

class State(TypedDict):
    messages: list[str]

# Example node calling an LLM (pseudo-code)
def llm_node(state: State) -> State:
    # Imagine calling an LLM here
    response = "AI response"
    return {"messages": state["messages"] + [response]}

graph = StateGraph(State)
graph.add_node("llm_node", llm_node)
graph.set_entry_point("llm_node")
graph.add_edge("llm_node", END)

# Use a memory saver for persistence
checkpointer = MemorySaver()
app = graph.compile(checkpointer=checkpointer)

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

output

['AI response']

Troubleshooting

If you see TypeError, ensure your state matches the TypedDict schema.
If app.invoke() hangs, check that all nodes have edges leading to END or other nodes.
For import errors, verify langgraph is installed in your environment.

✅

Key Takeaways

Define your agent state with Python TypedDict for type safety.
Use StateGraph to add nodes and edges representing agent logic.
Compile the graph and invoke it with initial state to run the agent.
Persist state with checkpointers for complex or long-running agents.
Integrate external APIs inside node functions for powerful AI agents.

Verified 2026-04

Verify ↗