Skip to content

Session Management

Session management in Strands Agents provides a robust mechanism for persisting agent state and conversation history across multiple interactions. This enables agents to maintain context and continuity even when the application restarts or when deployed in distributed environments.

A session represents all of stateful information that is needed by agents and multi-agent systems to function, including:

Single Agent Sessions:

  • Conversation history (messages)
  • Agent state (key-value storage)
  • Other stateful information (like Conversation Manager)

Multi-Agent Sessions:

  • Orchestrator state and configuration
  • Individual agent states and result within the orchestrator
  • Cross-agent shared state and context
  • Execution flow and node transition history

Strands provides built-in session persistence capabilities that automatically capture and restore this information, allowing agents to seamlessly continue conversations where they left off.

Beyond the built-in options, third-party session managers provide additional storage and memory capabilities.

Simply create an agent with a session manager and use it:

from strands import Agent
from strands.session import SnapshotSessionManager
from strands.storage import LocalFileStorage
# Create a snapshot session manager with a unique session ID
session_manager = SnapshotSessionManager(
session_id="test-session",
storage=LocalFileStorage("./sessions/"),
)
# Create an agent with the session manager
agent = Agent(session_manager=session_manager)
# Use the agent - all messages and state are automatically persisted
agent("Hello!") # This conversation is persisted

The conversation, and associated state, is persisted to the underlying storage backend.

FileSessionManager and S3SessionManager remain supported in Python, but use SnapshotSessionManager for new single-agent sessions.

Multi-agent systems (Graph/Swarm) can also use session management to persist their state.

from strands import Agent
from strands.multiagent import GraphBuilder
from strands.session import FileSessionManager
# Create agents
agent1 = Agent(name="researcher")
agent2 = Agent(name="writer")
# Create a repository-based session manager for the graph
session_manager = FileSessionManager(session_id="multi-agent-session")
# Build the graph with session management on the orchestrator
builder = GraphBuilder()
builder.add_node(agent1, "researcher")
builder.add_node(agent2, "writer")
builder.add_edge("researcher", "writer")
builder.set_session_manager(session_manager)
graph = builder.build()
result = graph("Research and write about AI")

Multi-agent session managers only track the current state of the Graph/Swarm execution and do not persist individual agent conversation histories.

Snapshot-based session managers accept any Storage backend, including InMemoryStorage, LocalFileStorage, S3Storage, and custom implementations. See Storage for backend configuration, tradeoffs, custom backends, and required S3 permissions.

Pass storage to SnapshotSessionManager or to the agent. Manager-level storage takes precedence over agent-level storage. If neither provides storage, SnapshotSessionManager uses LocalFileStorage("./.strands/").

Repository-based managers configure storage directly through FileSessionManager, S3SessionManager, or a custom SessionRepository. They do not use the unified Storage backend.

SnapshotSessionManager in Python and SessionManager in TypeScript persist a complete point-in-time snapshot. Both restore snapshot_latest during initialization and support immutable checkpoints.

Single Agent Events

  • Agent Initialization: Restores state from snapshot_latest if it exists.
  • Message Addition (save_latest_on="message"): Saves after each message and again when the invocation ends.
  • Agent Invocation (save_latest_on="invocation", default): Saves when the invocation ends.
  • Snapshot Trigger: Creates an immutable checkpoint when snapshot_trigger returns True.
  • Message Redaction: Flushes redacted content to the latest snapshot under every strategy.

Multi-Agent Events

SnapshotSessionManager does not support Graph or Swarm. Use a repository-based manager for Python multi-agent persistence.

FileSessionManager, S3SessionManager, and RepositorySessionManager store individual message records and agent metadata. Use them for existing repository-format sessions, Graph, Swarm, or bidirectional streaming.

SnapshotSessionManager remains the recommended option for new single-agent sessions. Use the repository-based managers when you need one of the compatibility cases above:

from strands import Agent
from strands.session import FileSessionManager, S3SessionManager
file_session_manager = FileSessionManager(
session_id="file-session",
storage_dir="./sessions/",
)
file_agent = Agent(session_manager=file_session_manager)
s3_session_manager = S3SessionManager(
session_id="s3-session",
bucket="my-agent-sessions",
prefix="production/",
)
s3_agent = Agent(session_manager=s3_session_manager)
  • Agent Initialization: Restores stored messages, agent state, conversation manager state, interrupt state, and model state.
  • Message Addition: Appends the message record and synchronizes changed agent state.
  • Agent Invocation: Synchronizes changed agent and conversation manager state.
  • Message Redaction: Updates the latest stored message record.
  • Multi-Agent and Bidirectional Events: Restore state during initialization and synchronize state after node, message, and invocation events.

In addition to snapshot_latest, snapshot-based session managers support immutable snapshots. Strands assigns each append-only checkpoint a UUIDv7 identifier, which lets you restore the agent to any prior checkpoint instead of only the latest state.

Use the snapshot trigger callback to control when an immutable snapshot is created. The callback receives the current agent data and returns a boolean:

from strands import Agent
from strands.session import SnapshotSessionManager
from strands.storage import LocalFileStorage
session_manager = SnapshotSessionManager(
session_id="my-session",
storage=LocalFileStorage("./sessions/"),
snapshot_trigger=lambda *, agent_data, **_: len(agent_data.messages) % 4 == 0,
)
agent = Agent(session_manager=session_manager)

Snapshot IDs are UUID v7, so they sort lexicographically in chronological order. Use list_snapshot_idslistSnapshotIds on the session manager to retrieve them, then pass a snapshot_idsnapshotId to restore_snapshotrestoreSnapshot:

import asyncio
from strands import Agent
from strands.session import SnapshotSessionManager
from strands.storage import LocalFileStorage
session_manager = SnapshotSessionManager(
session_id="my-session",
storage=LocalFileStorage("./sessions/"),
)
agent = Agent(session_manager=session_manager)
async def restore_snapshot() -> None:
snapshot_id = await session_manager.save_snapshot(agent, is_latest=False)
assert snapshot_id is not None
snapshot_ids = await session_manager.list_snapshot_ids(agent)
assert snapshot_id in snapshot_ids
await session_manager.restore_snapshot(agent, snapshot_id=snapshot_id)
asyncio.run(restore_snapshot())

To remove all snapshots for a session, call the session manager’s delete method:

import asyncio
from strands.session import SnapshotSessionManager
from strands.storage import LocalFileStorage
session_manager = SnapshotSessionManager(
session_id="my-session",
storage=LocalFileStorage("./sessions/"),
)
asyncio.run(session_manager.delete_session())

SnapshotSessionManager stores a versioned Snapshot JSON object. Its data field contains managed messages, agent state, conversation manager state, interrupt state, model state, and the system prompt.

The following record-based models apply to FileSessionManager, S3SessionManager, and RepositorySessionManager:

Session

The Session model is the top-level container for session data:

  • Purpose: Provides a namespace for organizing multiple agents and their interactions
  • Key Fields:
    • session_id: Unique identifier for the session
    • session_type: Type of session (currently "AGENT" for both agent & multiagent in order to keep backward compatibility)
    • created_at: ISO format timestamp of when the session was created
    • updated_at: ISO format timestamp of when the session was last updated

SessionAgent

The SessionAgent model stores agent-specific data:

  • Purpose: Maintains the state and configuration of a specific agent within a session
  • Key Fields:
    • agent_id: Unique identifier for the agent within the session
    • state: Dictionary containing the agent’s state data (key-value pairs)
    • conversation_manager_state: Dictionary containing the state of the conversation manager
    • created_at: ISO format timestamp of when the agent was created
    • updated_at: ISO format timestamp of when the agent was last updated

SessionMessage

The SessionMessage model stores individual messages in the conversation:

  • Purpose: Preserves the conversation history with support for message redaction
  • Key Fields:
    • message: The original message content (role, content blocks)
    • redact_message: Optional redacted version of the message (used when sensitive information is detected)
    • message_id: Index of the message in the agent’s messages array
    • created_at: ISO format timestamp of when the message was created
    • updated_at: ISO format timestamp of when the message was last updated

These data models work together to provide a complete representation of an agent’s state and conversation history. The session management system handles serialization and deserialization of these models, including special handling for binary data using base64 encoding.

Multi-Agent State

Multi-agent systems serialize their state as JSON objects containing:

  • Orchestrator Configuration: Settings, parameters, and execution preferences
  • Node State: Current execution state and node transition history
  • Shared Context: Cross-agent shared state and variables

The following third-party session managers extend Strands with additional storage and memory capabilities:

Session ManagerProviderDescriptionDocumentation
AgentCoreMemorySessionManagerAmazonAdvanced memory with intelligent retrieval using Amazon Bedrock AgentCore Memory. Supports both short-term memory (STM) and long-term memory (LTM) with strategies for user preferences, facts, and session summaries.View Documentation
Contribute Your OwnCommunityHave you built a session manager? Share it with the community!Learn How

For advanced use cases, you can implement your own session storage backend.

For new single-agent sessions, implement the unified Storage protocol and pass it to SnapshotSessionManager.

For repository-based sessions, create a custom session repository by implementing the SessionRepository interface:

from typing import Optional
from strands import Agent
from strands.session.repository_session_manager import RepositorySessionManager
from strands.session.session_repository import SessionRepository
from strands.types.session import Session, SessionAgent, SessionMessage
class CustomSessionRepository(SessionRepository):
"""Custom session repository implementation."""
def __init__(self):
"""Initialize with your custom storage backend."""
# Initialize your storage backend (e.g., database connection)
self.db = YourDatabaseClient()
def create_session(self, session: Session) -> Session:
"""Create a new session."""
self.db.sessions.insert(asdict(session))
return session
def read_session(self, session_id: str) -> Optional[Session]:
"""Read a session by ID."""
data = self.db.sessions.find_one({"session_id": session_id})
if data:
return Session.from_dict(data)
return None
# Implement other required methods...
# create_agent, read_agent, update_agent
# create_message, read_message, update_message, list_messages
# Use your custom repository with RepositorySessionManager
custom_repo = CustomSessionRepository()
session_manager = RepositorySessionManager(
session_id="user-789",
session_repository=custom_repo
)
agent = Agent(session_manager=session_manager)

This approach allows you to store session data in any backend system while leveraging the built-in session management logic.

Both file and S3 backends use the same key structure:

Snapshot-based layout (recommended for new single-agent sessions):

<root>/
└── session/
└── <session_id>/
└── scopes/
└── agent/
└── <agent_id>/
└── snapshots/
├── snapshot_latest.json
└── immutable_history/
└── snapshot_<uuid7>.json

Repository-based layout:

<root>/
└── session_<session_id>/
├── session.json
├── agents/
│ └── agent_<agent_id>/
│ ├── agent.json
│ └── messages/
│ ├── message_0.json
│ └── message_1.json
└── multi_agents/
└── multi_agent_<orchestrator_id>/
└── multi_agent.json

Using the same session ID and storage location does not migrate repository-based Python data to SnapshotSessionManager. Existing sessions can continue using their current manager, or applications can migrate the required state explicitly.

Session management is designed around a single live writer per conversation: the session ID plus the agent ID (agent_idid), or the orchestrator ID for a Graph or Swarm, address one conversation thread in storage.

Give each conversation its own session ID. Several agents can share one session ID as long as their agent IDs differ: the session acts as a namespace, and each agent keeps its own messages and state inside it. A single agent instance processes one invocation at a time by default and rejects overlap, as described in Concurrent Invocations.

Constructing an agent is cheap: it wires up tools, hooks, and plugins locally and makes no model call. Build one per request, invoke it, and let it go out of scope. The model provider is the part worth reusing: providers such as BedrockModel build their client in the constructor, so create the provider once per process and pass the same instance to every agent.

The built-in session managers take no distributed lock, and the single-instance invocation guard is in-process, so neither can detect a second writer running elsewhere. Two patterns result:

  • Two live agents addressing the same session ID and agent ID. The default IDs (agent_id="default"id: 'agent') make this easy to do by accident, including across separate executions that each build their own agent. Overlapping invocations overwrite each other’s turns, sequential ones merge two conversations into one history, and neither call errors.
  • Two callers creating the same session at the same time. Session creation is a check followed by a write, not an atomic operation, so simultaneous cold starts on a new session ID can both succeed, with the later write winning.

When implementing session persistence in your applications, consider these best practices:

  • Use Unique Session IDs: Generate unique session IDs for each user or conversation context to prevent data overlap.
  • Session Cleanup: Implement a strategy for cleaning up old or inactive sessions. Consider adding TTL (Time To Live) for sessions in production environments.
  • Understand Persistence Triggers: Remember that changes to agent state or messages are only persisted during specific lifecycle events.
  • Concurrent Access: Session managers are not thread-safe and take no distributed lock. See Sessions, Agents, and Concurrency.
  • Secure Storage Directories: The session storage directory is a trusted data store. Restrict filesystem permissions so that only the agent process can read and write to it. In shared or multi-tenant environments (shared volumes, containers), be aware that the SDK does not block symlinks in the session storage directory. If an attacker with write access to the storage directory creates a symlink (e.g., message_0.json pointing to an arbitrary file), the SDK will follow it, which could cause sensitive file contents to be loaded into the agent’s conversation history.