raydenai/queen-protocol

Queen Protocol is an operational contract for orchestrating parallel coding "colonies" (6–12 concurrent shards) on a single host. It coordinates Claude Code "queen" sessions, tmux-pane child sessions, Kimi background tasks in git worktrees, OpenAI Codex/Anthropic agent sidecars, and operator-invoked commands to dispatch, converge, verify and land changes across shards and repos. Who uses it: operators and teams running multi-model coding agents and automated worker tasks for real-code edits, cross-repo dispatch, and disciplined verification. Key features: - End-to-end pipeline: SURVEY → PLAN → DISPATCH → WATCH → CONVERGE → VERIFY → LAND - Polymorphic workers (queen, kimi, codex/agent sidecars, mesh panes) and per-shard routing - Enforced controls: concurrent-queen lock, atomic active.json writes, six-step report validation, files_allowed gate, integration-worktree snapshot/rollback, semantic-injection defenses, telemetry/metrics, secrets scanning - Commands for auto-converge, abort, versioning, and decompose/dispatch; hardened error/recovery paths - Single-host production-ish today; multi-host fencing and distributed locks deferred to v3.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.