Agent with Memory

Memory lets your agent remember facts about users across conversations. Unlike storage (which persists conversation history), memory stores user-level information like preferences and context.

Create a Python file

agent_with_memory.py

from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.memory import MemoryManager
from agno.models.openai import OpenAIResponses
from agno.tools.yfinance import YFinanceTools
from rich.pretty import pprint

db = SqliteDb(db_file="tmp/agents.db")

memory_manager = MemoryManager(
    model=OpenAIResponses(id="gpt-5.2"),
    db=db,
)

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    tools=[YFinanceTools(all=True)],
    db=db,
    memory_manager=memory_manager,
    enable_agentic_memory=True,
    markdown=True,
)

user_id = "investor@example.com"

# Tell the agent about yourself
agent.print_response(
    "I'm interested in AI and semiconductor stocks. My risk tolerance is moderate.",
    user_id=user_id,
    stream=True,
)

# Ask a follow-up using any preferences the agent chose to save
agent.print_response(
    "What stocks would you recommend for me?",
    user_id=user_id,
    stream=True,
)

# View stored memories
memories = agent.get_user_memories(user_id=user_id)
print("\nStored Memories:")
pprint(memories)

Set up your virtual environment

uv venv --python 3.12
source .venv/bin/activate

Install dependencies

uv pip install -U agno openai yfinance sqlalchemy rich

Export your OpenAI API key

export OPENAI_API_KEY="your_openai_api_key_here"

Run Agent

python agent_with_memory.py

Memory vs Storage

Feature	Storage	Memory
What it stores	Conversation history	User preferences and facts
Scope	Per session	Per user (across all sessions)
Use case	"What did we discuss?"	"What do you know about me?"

Enabling Memory

enable_agentic_memory=True (used above): Adds an update_user_memory tool. The model decides whether to call it. Existing memories are added to context.
update_memory_on_run=True: Runs the memory manager for each non-empty user input. This adds a model call and does not guarantee that every detail becomes a memory.