What problem does it solve?
Manually coordinating multiple parallel AI agent sessions across different harnesses like Claude Code and Codex is time-consuming and error-prone when splitting complex development tasks, leading to duplicated work and merge conflicts.
Core Features & Use Cases
- Multi-Harness Support: Run parallel agent sessions across Claude Code, Codex, OpenCode, Cline, Gemini, and Qwen in isolated tmux panes.
- Prebuilt Workflow Patterns: Access ready-to-use templates for common multi-agent tasks including research-and-implement splits, parallel multi-file feature development, test-fix loops, and cross-perspective code reviews.
- Conflict Avoidance: Integrate with git worktrees to isolate agent work on separate branches, eliminating file merge conflicts when combining parallel outputs.
- Use Case: Split a new billing feature build into three parallel tracks: one agent builds the database schema, a second builds the API endpoints, and a third builds the UI components, cutting total development time significantly.
Quick Start
Use the dmux-workflows skill to split your complex feature development task into parallel agent sessions across your preferred AI harnesses to accelerate delivery.