max-subagents-parallel-help

What problem does it solve?

Decomposes complex tasks into parallel subagents and orchestrates their execution using configurable waves, ensuring efficient exploration, task decomposition, and synthesis.

Core Features & Use Cases

• Flags: --agents, --depth, --ensemble, --strategy control agent count, recursion depth, ensemble mode, and execution method.
• Strategies: recursive (decompose -> parallel waves -> synthesize), ensemble (multiple models for same subtask -> compare & merge), sweep (breadth-first exploration with many angles).
• Model Tiering: assigns tasks to suitable models (haiku/sonnet/opus) based on type (finding, reading, architecture) to balance speed and depth.
• Execution Flow: parse flags, choose strategy, decompose into subtasks, group into waves, execute in order, mandatory synthesis to merge results.

Quick Start

Run max-subagents-parallel on a task with default settings to decompose the work and execute in parallel.