orchestrating-tmux-claudes

Coordinate multiple AI agents in tmux by launching a multi-AI session with Claude Code, codex, ollama, or haiku instances across separate panes, then orchestrate their workflows from a coordinator instance that monitors outputs, verifies quality gates, and issues commands with guaranteed execution via Enter as a separate argument.

Yes. Multi-agent orchestration automates complex workflows by delegating specialized tasks—test writing, implementation, refactoring, code review—to separate AI agents within a single tmux session, with the coordinator handling real-time monitoring, auto-recovery, and role-based task routing.

Prerequisites include tmux, Node.js, Claude Code CLI, and Git. The orchestration reads pane outputs before decisions, logs breadcrumbs to /tmp, and executes actual bash commands without hallucination, requiring these tools installed and Git initialized in your project.

Enforce role-based skill loading by restricting the coordinator role to the main Claude instance only, preventing worker agents from attempting self-orchestration and ensuring clean separation of concerns across the tmux session.

Auto-recovery monitoring detects stuck agents and automatically recovers them, ensuring continuous progress through the workflow and reducing manual intervention overhead in long-running multi-stage development processes.

Yes. Tmux orchestration coordinates AI agents in either parallel or sequential workflows within separate panes, with the coordinator deciding execution order and managing inter-agent dependencies based on quality gates and verification outcomes.