What is stateless and stateful in AI agents?

Stateless vs Stateful in AI Agents

State management is a critical design decision when building AI agents. Understanding stateless vs stateful architecture determines how your agent handles context, memory, and multi-turn conversations.

Core Definitions

Stateless Agent

pythonCopy
# Stateless: each call has no memory of previous calls
def stateless_agent(user_message: str) -> str:
    response = client.messages.create(
        model="claude-opus-4-6",
        messages=[
            {"role": "user", "content": user_message}
        ]
    )
    return response.content[0].text

# Call 1
reply1 = stateless_agent("My name is Alice")
# Call 2 — agent has NO idea Alice mentioned her name
reply2 = stateless_agent("What is my name?")  # Will say "I don't know"

Stateful Agent

pythonCopy
# Stateful: messages accumulate in history
class StatefulAgent:
    def __init__(self):
        self.history = []  # Maintains state

    def chat(self, user_message: str) -> str:
        self.history.append({"role": "user", "content": user_message})

        response = client.messages.create(
            model="claude-opus-4-6",
            messages=self.history
        )

        assistant_msg = response.content[0].text
        self.history.append({"role": "assistant", "content": assistant_msg})
        return assistant_msg

agent = StatefulAgent()
agent.chat("My name is Alice")   # Stored in history
agent.chat("What is my name?")   # Returns "Alice" — context retained

LangGraph Stateful Agents

LangGraph manages state explicitly via a

State

pythonCopy
from langgraph.graph import StateGraph
from typing import TypedDict, Annotated
from langgraph.graph.message import add_messages

class State(TypedDict):
    messages: Annotated[list, add_messages]  # State persisted across nodes
    current_task: str
    completed_steps: list

graph = StateGraph(State)
# State flows through all nodes — each node reads and updates it

Types of State in Agents

When to Use Each

Key Challenge: State Management at Scale

Stateful agents are more powerful but require careful design:

• Where to store state (in-memory vs external DB)
• How to handle concurrency (multiple users)
• When to expire/summarize old state (token limits)
• How to recover from failures (checkpointing)