ralph-eats-beads

You can automate an autonomous coding loop by orchestrating Claude agents that pull tasks from bd ready and execute them in parallel tmux subwindows. This allows multiple subagents to work on different issues simultaneously, improving throughput.

To set up an autonomous AI loop, initialize bd with an epic and initial issues, generate a plan with scripts/planner.sh, and start the loop with scripts/ralph.sh. The spawned subagents automatically sync progress and close bd tasks upon successful implementation.

Yes, running parallel AI agents requires the bd tool, tmux, and Claude CLI. These dependencies are essential because bd provides the tasks, tmux manages parallel subwindows, and Claude CLI executes the coding work.

The best way to monitor parallel AI agents is using tmux windows for real-time visibility into each subagent's progress. Additionally, a watchdog script monitors spawned subagents to handle stalled work and ensure continuous progress across all active tasks.

The watchdog handles stalled work by monitoring spawned Claude subagents in tmux and intervening to ensure progress. It synchronizes parallel work across subwindows to prevent bottlenecks and keep the autonomous coding loop running continuously.

Yes, you can use Claude CLI for plan-driven development with bd epics and issues. The autonomous coding loop aligns subagents with bd tasks, automatically syncing progress and closing issues when implementation succeeds.