Show HN: Pi-cowboy – use Codex, Claude Code, OpenCode as subagents (github.com)

The launch of Pi-cowboy marks a significant advancement in task orchestration for AI agents. As a multi-agent orchestrator, it enables users to divide complex jobs into smaller, more manageable subtasks that can be executed in parallel using various AI models such as Codex, Claude Code, and OpenCode. Each agent operates within its own terminal pane in a Herdr workspace, allowing for real-time monitoring, interaction, and streamlined management of tasks. Notably, the use of copy-on-write Git worktrees allows for efficient disk usage and high parallelism—even with extensive repositories. For the AI/ML community, Pi-cowboy represents a leap in operational efficiency, as it minimizes errors and complexity by assigning clear, isolated tasks. With built-in features for task templates and customizable concurrency settings, users can create tailored workflows that suit their project needs, whether for code audits or generating release notes. This innovation not only enhances productivity but also ensures a cleaner, safer approach to version control and task management, highlighting the increasing integration of AI tools in software development workflows.