What problem does it solve?
Wave-parallelism provides a methodology for safely executing independent tasks in parallel using subagent waves. Tasks are grouped into waves based on dependency analysis — all tasks in a wave can run simultaneously, waves execute sequentially.
Two responsibilities:
- Safety — dependency analysis, file-ownership, git contention handling ensure parallel agents don't conflict.
- Orchestration — wave assignment, completion verification, inter-wave validation ensure correct execution order and results.
Core Features & Use Cases
- Dependency analysis outlines needs, creates, files_modified for each task to build the dependency graph.
- File ownership safety ensures one file is owned by a single agent per wave to prevent merge conflicts.
- Wave scheduling & verification sequences waves and validates completion before proceeding.
- Use Case: Run multiple feature studies in parallel to explore different paths, then orchestrate safe merges.
Quick Start
To begin, define tasks with clear dependencies and let the orchestrator dispatch waves in the current project context.