Show and tell: ORCH — CLI orchestrator for running multiple Tabby agents as a coordinated team

[oxgeneral]

Hey Tabby community!

Sharing ORCH — an open-source CLI runtime that coordinates teams of AI coding agents: https://github.com/oxgeneral/ORCH

Tabby provides the self-hosted coding assistant. ORCH sits on top: it coordinates multiple Tabby (or other AI) agents working on the same codebase simultaneously, solving the multi-agent coordination problem.

What ORCH does:

• Typed task queue — state machine todo → in_progress → review → done, with automatic retry on stalls/failures
• Scope locking — prevents two agents from editing the same file at the same time
• Inter-agent messaging — agents pass results to each other without human copy-paste
• Adapter system — shell adapter works with any coding tool, including Tabby-backed workflows
• TUI dashboard — real-time terminal view of all running agents, tasks, and logs
• Goals system — decompose high-level objectives into agent subtasks

Quick start:

npm install -g @oxgeneral/orch
orch agent add "Agent" --adapter shell  # shell adapter works with any tool
orch task add "Refactor API layer" --assignee <agent-id>
orch tui

1,647 passing tests, TypeScript strict, zero-database (YAML/JSON files), MIT.
https://github.com/oxgeneral/ORCH

Curious if Tabby users are running multi-agent setups — what coordination patterns work for you?