Entity Memory Store

The Entity Memory Store captures structured knowledge about external entities - companies, people, projects, and systems. Think of it as your agent’s professional “mental rolodex” that accumulates knowledge over time.

Overview

Aspect	Details
Data Type	Entities with facts, events, and relationships
Scope	Configurable (global or per-user)
Persistence	Long-term
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, EntityMemoryConfig
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(
        entity_memory=True,
    ),
)

# Entities are extracted automatically
agent.print_response(
    "Just met with Acme Corp. They're a fintech startup in SF, "
    "50 employees. CTO is Jane Smith. They use Python and Postgres.",
    user_id="[email protected]",
    session_id="session_1",
)

# Later, entity knowledge is recalled
agent.print_response(
    "What do we know about Acme Corp?",
    user_id="[email protected]",
    session_id="session_2",
)

Three Types of Knowledge

Entity Memory captures three types of information:

Facts (Timeless Truths)

Facts are things that remain true over time:

“Uses PostgreSQL for their main database”
“Headquarters in San Francisco”
“50 employees”
“Founded by ex-Google engineers”

Events (Time-Bound Occurrences)

Events are things that happened at a specific time:

“Launched v2.0 on January 15, 2025”
“Closed $50M Series B led by Sequoia”
“Had 4-hour outage affecting payment processing”
“Announced partnership with CloudCo yesterday”

Relationships (Entity Connections)

Relationships link entities together:

Jane Smith → CEO → Acme Corp
Project Atlas → uses → PostgreSQL
Acme Corp → competitor_of → Beta Inc

ALWAYS Mode (Default)

In ALWAYS mode, entities are extracted automatically from conversations:

from agno.learn import LearningMachine, LearningMode, EntityMemoryConfig

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

Behavior:

Extraction happens in parallel with the agent’s response
No tools are visible to the agent
All conversations are analyzed for entity information

AGENTIC Mode

In AGENTIC mode, the agent explicitly decides when to create or update entities:

from agno.learn import LearningMachine, LearningMode, EntityMemoryConfig

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

# Agent uses tools to manage entities
agent.print_response(
    "Create an entry for Acme Corp - they're a fintech startup with 50 employees.",
    user_id="[email protected]",
)

Available tools:

Tool	Description
`search_entities`	Search for entities by name, facts, or events
`create_entity`	Create a new entity
`update_entity`	Update entity properties
`add_fact`	Add a timeless truth
`update_fact`	Correct an existing fact
`delete_fact`	Remove an incorrect fact
`add_event`	Record a time-bound occurrence
`add_relationship`	Create connection between entities

Entity Data Model

Each entity contains:

Field	Type	Description
`entity_id`	`str`	Unique identifier (e.g., “acme_corp”)
`entity_type`	`str`	Category: “company”, “person”, “project”, etc.
`name`	`str`	Display name
`description`	`str`	Brief description
`properties`	`dict`	Key-value metadata
`facts`	`list`	Timeless truths
`events`	`list`	Time-bound occurrences
`relationships`	`list`	Connections to other entities

Accessing Entity Memory

# Get the learning machine
learning = agent.get_learning_machine()

# Search for entities
entities = learning.entity_memory_store.search(
    query="acme",
    entity_type="company",
    limit=10
)

for entity in entities:
    print(f"Entity: {entity.name}")
    print(f"Facts: {entity.facts}")
    print(f"Events: {entity.events}")
    print(f"Relationships: {entity.relationships}")

# Debug output
learning.entity_memory_store.print(entity_id="acme_corp", entity_type="company")

Context Injection

Relevant entities are injected into the agent’s context:

<entity_memory>
**Known information about relevant entities:**

**Acme Corp** (company)
Properties: industry: fintech, size: 50 employees

Facts:
  - Uses PostgreSQL and Redis for their data layer
  - Headquarters in San Francisco

Events:
  - Launched v2.0 with new ML features (2025-01-15)
  - Closed $50M Series B led by Sequoia (2024-Q3)

Relationships:
  - CEO: jane_smith
  - competitor_of: beta_inc
</entity_memory>

Namespace Configuration

Control who can access entity data:

from agno.learn import LearningMachine, EntityMemoryConfig

# Global: shared with everyone (default)
entity_memory=EntityMemoryConfig(namespace="global")

# User: private per user
entity_memory=EntityMemoryConfig(namespace="user")

# Custom: explicit grouping
entity_memory=EntityMemoryConfig(namespace="sales_team")

Example: CRM Assistant

from agno.agent import Agent
from agno.db.postgres import PostgresDb
from agno.learn import LearningMachine, LearningMode, EntityMemoryConfig
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(
        entity_memory=EntityMemoryConfig(
            mode=LearningMode.AGENTIC,
            namespace="sales_team",
        ),
    ),
    instructions="""You are a CRM assistant. Track information about companies
    and contacts. Always search for existing entities before creating new ones.
    Record important facts, events, and relationships.""",
)

# Track a new company
agent.print_response(
    "Just had a great call with TechStartup Inc. They're Series A, "
    "20 employees, looking for data infrastructure. CEO is Bob Johnson, "
    "he was previously at Google. They're planning to close their round next month.",
    user_id="[email protected]",
    session_id="session_1",
)

# Prepare for follow-up
agent.print_response(
    "I have a follow-up call with TechStartup tomorrow. What do we know about them?",
    user_id="[email protected]",
    session_id="session_2",
)

Facts vs Events Guide

Use Facts For	Use Events For
Tech stack	Product launches
Headquarters location	Funding rounds
Employee count	Outages or incidents
Industry/domain	Partnerships announced
Pricing model	Key meetings
Preferences	Milestones achieved

Building Relationship Graphs

Create rich knowledge graphs by linking entities:

# The agent can build relationships like:
# Jane Smith --[CEO]--> Acme Corp
# Jane Smith --[former_employee]--> Google
# Acme Corp --[competitor_of]--> Beta Inc
# Project Atlas --[uses]--> PostgreSQL

Common relationship types:

People: CEO, CTO, engineer_at, founder, reports_to
Companies: competitor_of, partner_of, acquired_by, subsidiary_of
Projects: uses, depends_on, integrates_with, owned_by

Get Started

Basics

Context Management

Execution Control

Additional Features

Integrations

Other

Entity Memory Store

Overview

Basic Usage

Three Types of Knowledge

Facts (Timeless Truths)

Events (Time-Bound Occurrences)

Relationships (Entity Connections)

ALWAYS Mode (Default)

AGENTIC Mode

Entity Data Model

Accessing Entity Memory

Context Injection

Namespace Configuration

Example: CRM Assistant

Facts vs Events Guide

Building Relationship Graphs

Get Started

Basics

Context Management

Execution Control

Additional Features

Integrations

Other

​Overview

​Basic Usage

​Three Types of Knowledge

​Facts (Timeless Truths)

​Events (Time-Bound Occurrences)

​Relationships (Entity Connections)

​ALWAYS Mode (Default)

​AGENTIC Mode

​Entity Data Model

​Accessing Entity Memory

​Context Injection

​Namespace Configuration

​Example: CRM Assistant

​Facts vs Events Guide

​Building Relationship Graphs

Overview

Basic Usage

Three Types of Knowledge

Facts (Timeless Truths)

Events (Time-Bound Occurrences)

Relationships (Entity Connections)

ALWAYS Mode (Default)

AGENTIC Mode

Entity Data Model

Accessing Entity Memory

Context Injection

Namespace Configuration

Example: CRM Assistant

Facts vs Events Guide

Building Relationship Graphs