Learning

Agents that learn, adapt, and improve over time using the Learning Machine.

The Learning Machine enables agents to learn from every interaction. Instead of building memory, knowledge, and feedback systems separately, configure one system that handles all learning.

The goal: An agent on interaction 1000 is fundamentally better than it was on interaction 1.

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

db = PostgresDb(db_url="postgresql+psycopg://ai:ai@localhost:5532/ai")

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    db=db,
    learning=True,  # Enable all learning stores
)

# Session 1
agent.run("I'm Alex, I prefer concise answers.", user_id="[e-mail]")

# Session 2 - agent remembers
agent.run("What do you know about me?", user_id="[e-mail]")
# -> "You're Alex, you prefer concise answers"

Learning Stores

The Learning Machine coordinates five specialized stores:

Store	What It Captures	Scope
User Profile	Structured fields (name, preferences)	Per user
User Memory	Unstructured observations	Per user
Session Context	Goal, plan, progress, summary	Per session
Entity Memory	Facts, events, relationships	Configurable
Learned Knowledge	Insights, patterns, best practices	Global

Configuration Levels

from agno.learn import LearningMachine, UserProfileConfig, SessionContextConfig, LearningMode

# Level 1: Dead Simple
agent = Agent(model=model, db=db, learning=True)

# Level 2: Pick What You Want
agent = Agent(
    model=model,
    db=db,
    learning=LearningMachine(
        user_profile=True,
        session_context=True,
        entity_memory=False,
        learned_knowledge=False,
    ),
)

# Level 3: Full Control
agent = Agent(
    model=model,
    db=db,
    learning=LearningMachine(
        user_profile=UserProfileConfig(mode=LearningMode.AGENTIC),
        session_context=SessionContextConfig(enable_planning=True),
    ),
)

Learning Modes

Each store can run in different modes:

Mode	Behavior	Best For
ALWAYS	Automatic extraction after each turn	User profiles, session context
AGENTIC	Agent decides when to save	Learned knowledge
PROPOSE	Agent proposes, user confirms	High-stakes knowledge

Featured Patterns

Personal Assistant

Remembers preferences, tracks context, learns from feedback.

Support Agent

Multi-tenant learning with namespace isolation. Tracks customer context and accumulates support knowledge.

Run the Examples

git clone https://github.com/agno-agi/agno.git
cd agno/cookbook/08_learning

# Setup
./setup_venv.sh

# Run examples
python 01_basics/1a_user_profile_always.py
python 07_patterns/personal_assistant.py