User Memory

Unstructured observations about users.

The User Memory Store captures unstructured observations about users: preferences, behaviors, and context that don’t fit into structured profile fields.

Aspect	Value
Scope	Per user
Persistence	Long-term (with optional curation)
Default mode	Always
Supported modes	Always, Agentic

Basic Usage

from agno.agent import Agent
from agno.db.postgres import PostgresDb
from agno.learn import LearningMachine
from agno.models.openai import OpenAIResponses

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    db=PostgresDb(db_url="postgresql+psycopg://ai:ai@localhost:5532/ai"),
    learning=LearningMachine(user_memory=True),
)

# Session 1: Share preferences
agent.print_response(
    "I prefer code examples over explanations. Also, I'm working on a machine learning project.",
    user_id="[e-mail]",
    session_id="session_1",
)

# Session 2: Memory is recalled
agent.print_response(
    "Explain async/await in Python",
    user_id="[e-mail]",
    session_id="session_2",
)

The agent knows to include code examples and may relate to ML context.

Always Mode

Memories are extracted automatically after each response.

from agno.learn import LearningMachine, LearningMode, UserMemoryConfig

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    db=db,
    learning=LearningMachine(
        user_memory=UserMemoryConfig(mode=LearningMode.ALWAYS),
    ),
)

Tradeoff: extra LLM call per interaction.

Agentic Mode

The agent receives tools to manage memories explicitly.

from agno.learn import LearningMachine, LearningMode, UserMemoryConfig

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    db=db,
    learning=LearningMachine(
        user_memory=UserMemoryConfig(mode=LearningMode.AGENTIC),
    ),
)

agent.print_response(
    "Remember that I always want to see error handling in code examples.",
    user_id="[e-mail]",
)

Available tool: update_user_memory (supports add, update, delete, and clear operations)

Tradeoff: agent may miss implicit observations.

What Gets Captured

Good for User Memory	Better for User Profile
”Prefers detailed explanations”	Name: “Alice Chen"
"Working on ML project”	Company: “Acme Corp"
"Struggles with async code”	Role: “Data Scientist"
"Uses VS Code”	Timezone: “PST”

Memory Data Model

Field	Description
user_id	User this memory belongs to
memories	List of memory entries (id, content, optional metadata)
agent_id	Agent context for audit trail
team_id	Team context for audit trail
created_at	When created
updated_at	Last update

Accessing Memories

lm = agent.get_learning_machine()

# Get all memories
memories = lm.user_memory_store.get(user_id="[e-mail]")
if memories:
    for memory in memories.memories:
        print(f"- {memory.get('content')}")

# Debug output
lm.user_memory_store.print(user_id="[e-mail]")

Context Injection

Relevant memories are injected into the system prompt:

<user_memory>
- Prefers code examples over explanations
- Working on a machine learning project
- Uses Python 3.11
- Prefers concise responses
</user_memory>

Curation

Over time, memories accumulate. Use the Curator to maintain them:

lm = agent.get_learning_machine()

# Remove memories older than 90 days
lm.curator.prune(user_id="[e-mail]", max_age_days=90)

# Remove duplicates
lm.curator.deduplicate(user_id="[e-mail]")

Combining with User Profile

Use both stores for comprehensive user understanding:

from agno.learn import LearningMachine

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    db=db,
    learning=LearningMachine(
        user_profile=True,  # Structured: name, company
        user_memory=True,   # Unstructured: preferences, context
    ),
)