Session Storage

Store and retrieve agent, team, and workflow sessions from a database.

Adding a database persists runs under a session_id. Agent and team sessions contain conversation runs. Workflow sessions contain workflow runs.

Prerequisites

For the SQLite/OpenAI examples:

uv pip install -U "agno[openai,sqlite]"
export OPENAI_API_KEY="your_openai_api_key"

Configure the Session Table

SQL adapters default to agno_sessions for session metadata and agno_runs for runs. Supported adapters create a required table on the first relevant operation when it does not exist. Set runs_table as well as session_table when separating environments.

Use session_table to store sessions in a custom table:

The PostgreSQL fragment requires a running PostgreSQL database and its connection URL. Install the driver with uv pip install -U "psycopg[binary]", then replace the sample credentials below.

from agno.agent import Agent
from agno.db.postgres import PostgresDb

db = PostgresDb(
    db_url="postgresql+psycopg://user:password@localhost:5432/mydb",
    session_table="my_agent_sessions",
)

agent = Agent(db=db)

Use separate tables when you need database-level isolation between environments.

What Gets Stored

The following describes a hydrated session object returned by get_session(). SQL adapters store run records separately in runs_table and load them into session.runs; runs is not a column in the physical session table. Fields depend on the owner type and may be absent or None. The physical SQL row also stores a session_type discriminator; it is not a session-object attribute.

FieldTypeDescription
session_idstrUnique session identifier
agent_idstrThe agent ID (if agent session)
team_idstrThe team ID (if team session)
workflow_idstrThe workflow ID (if workflow session)
user_idstrThe user this session belongs to
session_datadictSession-specific data and state
agent_datadictAgent configuration and metadata
team_datadictTeam configuration and metadata
workflow_datadictWorkflow configuration and metadata
metadatadictAdditional custom metadata
runslistAll the runs (interactions) in this session
summarySessionSummaryAgent/team summary when enabled; not a WorkflowSession field
created_atintUnix timestamp when session was created
updated_atintUnix timestamp of last update

Retrieve Sessions

Use get_session() to retrieve a stored session:

from agno.agent import Agent
from agno.db.sqlite import SqliteDb

agent = Agent(db=SqliteDb(db_file="agent.db"))

agent.print_response("What is the capital of France?", session_id="session_123")

# Retrieve the session
session = agent.get_session(session_id="session_123")

# Access session data
print(session.session_id)
print(session.runs)  # List of runs with messages and responses

Works With Teams and Workflows

Teams and workflows expose the same get_session() interface:

from agno.agent import Agent
from agno.team import Team
from agno.workflow import Workflow
from agno.db.sqlite import SqliteDb

db = SqliteDb(db_file="agno.db")

team = Team(members=[Agent(name="Assistant")], db=db)
workflow = Workflow(db=db)

# Retrieve sessions the same way
team_session = team.get_session(session_id="team_session_123")
workflow_session = workflow.get_session(session_id="workflow_session_456")

Workflow sessions store complete pipeline runs rather than conversation messages. See Workflow Sessions for details.

Next Steps

Session Summaries

Condense long conversations to save tokens.

Storage Control

Choose what gets persisted to your database.

Developer Resources