🔄

Multi-Worker System

Active

Coordination system for running multiple Claude Code instances in parallel. Features lock management, shared state, and policy-based automation.

Bash Docker Coordination

Worker Workflow

Worker reads all status files, policies, and backlog

Worker updates their status file with project name

Worker completes tasks following policies

Worker logs changes to CHANGELOG.md

Worker sets project to "-" when done

Prevents multiple workers from editing the same project simultaneously. Workers claim locks via their status files.

Common files for policies, backlog, and environment settings that all workers can read and follow.

Each worker maintains a status file showing current project, task, and recent completions.

Pre-approved actions that workers can take without asking. Boundaries that require explicit approval.

Workers use the cheapest capable model: Ollama (free) → Haiku (cheap) → Opus (full price).

Statusline shows model usage (O:X H:Y L:Z) to monitor costs across all workers.

WORKER1-4.md	Individual worker status and current task
LOCKS.md	Project lock registry and ownership
POLICIES.md	Pre-approved actions and boundaries
BACKLOG.md	Shared task queue and ideas
CLAUDE.md	Master instructions for all workers