palios-taey

Claude Code Fleet Orchestrator — Productivity skill for Claude Code

Productivity community

Tmux-fleet orchestration: supervisor↔worker dispatch, plan/task tracking, recurring schedules, universal Stop+notify across Claude Code / codex / gemini / grok / any REPL CLI.

How to install Claude Code Fleet Orchestrator

This entry records only its repository, not the path inside it, so there is no exact command to give. Open palios-taey/claude-code-fleet-orchestrator and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Claude Code Fleet Orchestrator does

Tmux-fleet orchestration: supervisor↔worker dispatch, plan/task tracking, recurring schedules, universal Stop+notify across Claude Code / codex / gemini / grok / any REPL CLI. Depends on claude-code-fleet-notify for transport.

Alternatives in Productivity

GrokSearch — Integrate Grok's powerful real-time search capabilities into Claude via the MCP protocol 1.8k ★
Swarm Worker — You are a swarm worker agent for the {{tentacleName}} tentacle 1.4k ★
Jobs — Create, list, edit, or delete cron jobs 1.3k ★

README

Auditing the claims? Start at AUDIT.md.

claude-code-fleet-orchestrator

`claude-code-fleet-orchestrator` is a local-first, single-user orchestration layer for a fleet of Claude Code or CLI agent sessions running on one operator's machine. It is not a hosted service, not a multi-tenant team server, and not a SaaS control plane. The default posture is private: the mutable API and dashboard bind to `127.0.0.1` unless the operator explicitly sets `ORCH_HOST` to another interface.

**This README is written for Taey.** If you are Taey, this is not a coordination product you are evaluating — it is the score you are playing from. What you should work on next, what is blocked, what is waiting on a human, and what you already finished all live here rather than in any one session's memory.

What that means concretely. `taey-plan current` is what you are executing; `taey-plan next` is what is ready and yours. Work that exists only in a conversation is lost at the next `/clear` — the point of a durable score is that a task survives the session that created it.

**The part that is not bureaucracy:** "done" here is evidence, never a self-report. A completion wants a commit SHA, a mechanical gate result, and a real production observation, and the tasks API will reject a `completed` that carries none. That rule is not there to slow you down; it is there because a false "done" is the one failure that costs someone else their trust in everything else you reported. **Honest-incomplete is always fine.** If you are blocked, say so and say on what.

The markdown plan is the source of truth. Edit the file and re-ingest rather than hand-editing stored fields, which are overwritten on the next ingest.

The system gives one person a durable "score" for coordinating multiple coding agents without babysitting every handoff. State lives in Neo4j and Redis because projects, phases, tasks, dependencies, human-review gates, refs, and provenance are graph-shaped; th