ORCH — CLI orchestrator for AI agent teams (Claude, Codex, Cursor)

[oxgeneral]

Hey Goose community! 👋

I built ORCH — a CLI runtime that orchestrates AI coding agents (Claude Code, OpenCode, Codex, Cursor) as a coordinated engineering team. If you use Goose for coding tasks, ORCH can act as the orchestration layer on top — managing multiple agents in parallel with structure.

What it does

Instead of manually coordinating agents, ORCH runs them through a formal state machine:

todo → in_progress → review → done

Each task goes through mandatory review before completion — no silent failures, no unchecked merges.

Key features

• State machine task queue — tasks flow through typed states with validated transitions
• Auto-retry — failed agents automatically enter retrying state with configurable attempts
• Inter-agent messaging — agents communicate directly via orch msg send
• Shared context store — agents share state with orch context set/get
• Goal → Task hierarchy — high-level goals decompose into typed task queues
• Teams — group agents by role (backend, QA, reviewer)
• TUI dashboard — live Ink/React terminal UI showing all agents and tasks
• 1,647 tests — TypeScript strict mode, 100% ESM
• MIT license

Quick start

npm install -g @oxgeneral/orch
orch agent list           # see pre-configured agents
orch task add "Build API" # create task in state machine
orch tui                  # open live dashboard

How it complements Goose

Goose excels at executing individual coding tasks autonomously. ORCH adds the coordination layer for teams of agents:

• Run Goose alongside Claude Code, Codex, or Cursor — each on different tasks
• Tasks are assigned, tracked, and retried automatically
• Agents can share context and message each other between tasks
• Review gate ensures no task completes without validation

GitHub: https://github.com/oxgeneral/ORCH
npm: npm install -g @oxgeneral/orch

Happy to answer questions about the architecture or how it pairs with Goose!