orchestrating-swarms

Multi-agent swarms coordinate parallel code reviews by assigning leader and teammate roles to agents, distributing tasks across parallel and sequential workflows, and routing status updates through an inbox message system. This enables dependency-aware pipeline execution and scalable software delivery.

Agent pipelines with dependencies support environment spawn backends for in-process execution, tmux, and iTerm2. You can configure these backends to manage parallel and sequential workflows, allowing teammates to operate concurrently within your preferred terminal environment.

Available orchestration patterns include Parallel Specialists, Pipeline, Swarm, and Research + Implementation. These patterns manage complex workflows by organizing self-organizing swarms that spawn, assign, and monitor tasks through leader-teammate hierarchies and inbox communication.

To start a parallel workflow, spawn a named team, create a set of tasks, and launch teammates to begin execution. The task system integration handles assignment and monitoring, while the TeammateTool operations enable agents to self-organize and coordinate via inbox messages.

Orchestrating swarms requires no external dependencies or components. It operates entirely self-contained, providing built-in TeammateTool operations, task system integration, message formats, and environment spawn backends for in-process, tmux, and iTerm2 execution.

Use swarm orchestration patterns instead of sequential execution when managing dependency-aware pipelines, parallel code reviews, or complex software delivery workflows. Swarms allow self-organizing agents to spawn, assign, and monitor tasks concurrently, significantly reducing execution time for parallelizable operations.