deep-agents-orchestration

You can coordinate subagents for complex multi-step planning by using orchestration middleware to delegate work to specialized subagents and tracking progress with a TodoList workflow.

Human-in-the-loop approval for agent automation is a workflow that optionally requires human oversight to validate sensitive actions before execution, ensuring safety checks and retry controls are applied during production-grade operations.

You set up a deep agent with subagents and planning workflows by configuring create_deep_agent() to automatically include SubAgentMiddleware for delegation and TodoListMiddleware for task tracking.

Yes, you can add human-in-the-loop interrupts to an agent automation pipeline by integrating HumanInTheLoopMiddleware to manage sensitive actions and enable approval workflows within your execution flow.

Orchestration for multi-step investigations is needed when complex tasks can be split, tracked, and validated before execution, requiring coordinated delegation and optional human oversight to handle sensitive decisions safely.