Agency Swarm conveniently manages the state of your agents and threads, simplifying the agent creation and management process.

​
Agent State Management

With the new https://openai.github.io/openai-agents-python, agents (formerly assistants) are now stateless - eliminating the need to save and pass agent settings for chat persistence. The framework only needs to maintain the conversation history, which Agency Swarm manages through its thread system.

​
Thread State Management

Since managing conversation state is now a responsibility of the developer in the Agents SDK, the Agency Swarm framework has introduced a thread management system to handle, automate, and simplify history management. All agent outputs are processed and stored locally to allow for seamless agency usage within a single session. To maintain persistence across multiple sessions, the framework also provides convenient callback methods that require minimal input to set up.

Automatic Context Management

Handles agents outputs and automatically converts them into Agents SDK - compatible chat history

Persistent Conversations

Maintains ongoing conversations without manual history management.

​
Next Steps