ORCH is an open-source multi-agent orchestration harness created by oxgeneral. It acts as a supervisory control layer for running, coordinating, and automating complex software engineering tasks across multiple terminal-based AI coding agents and backend model providers. In daily use, ORCH operates through CLI commands and an optional headless background daemon. Developers define goals and assign them to specific agent personas or adapters—such as pi-coding-agent, Grok, Antigravity, or Claude Code integration skills. ORCH communicates with agent sub-processes over persistent RPC channels, tracking task lifecycles, monitoring standard input/output, and recording token usage and file mutations into structured event logs (run.jsonl). Unlike standalone coding assistants that operate as single-turn chatbots, ORCH specializes in multi-agent dispatch and execution state management. It provides role-based goal assignment, process lifecycle oversight, and telemetry capture across heterogeneous agent backends. ORCH is open source and distributed as a Node.js package via npm and GitHub.
Multi-Agent Task Orchestration — Dispatches high-level engineering goals to distinct AI agents based on configured roles and assigned capabilities. - Agent Adapter Ecosystem — Interfaces with multiple agent runtimes including pi-coding-agent, Claude Code skills, Grok, and Antigravity adapters. - Persistent RPC Management — Maintains long-lived child process sessions over standard I/O RPC to orchestrate multi-turn agent runs. - Headless Daemon Service — Enables background task execution and unattended goal processing decoupled from the active terminal session. - Structured Event Logging — Captures step-by-step agent telemetry, modified files, and token consumption metrics into JSONL log files.
Use cases
Use Case: Delegated Multi-Agent Development A developer breaks a large refactoring task into distinct sub-tasks, assigns specific modules to different agent adapters (e.g., specialized test writers vs. core logic implementers), and lets ORCH coordinate execution across those roles.
Use Case: Unattended Background Task Execution A team runs ORCH in headless daemon mode to automatically pick up unassigned repository maintenance tasks or issue goals, executing code modifications and logging results without requiring an open interactive terminal.
Use Case: Coding Agent Auditing and Telemetry A team evaluates different AI agent backends by running tasks through ORCH to collect standardized token counts, execution turn traces, and file change metrics recorded in run.jsonl.
Who it is for
ORCH is designed for software engineers, systems developers, and AI researchers who want to orchestrate multi-agent workflows and automate complex programming tasks across different agent runtimes. It is not intended for developers seeking a simple inline autocomplete plugin or single-prompt interactive chat interface.
Install guide
Installation Install ORCH globally using npm: bash npm install -g orch Or clone the repository to run from source: bash git clone https://github.com/oxgeneral/ORCH.git cd ORCH npm install npm run build
First Run Verify that the CLI is accessible and inspect the available agent adapters and roles: bash orch agent list Ensure underlying agent dependencies (such as pi-coding-agent or provider API keys like OpenRouter/Anthropic) are configured in your environment.
Example Session Create a goal and assign it to a specific agent role: bash # Create a task assigned to a designated coding agent orch goal create "Implement unit tests for auth service" --assignee coder # Run the orchestrator daemon or execute pending goals orch run
amux — amux is an open-source agent harness and session multiplexer developed by Mixpeek. Built in Rust, it is designed to manage,…
Pi Boss — Pi Boss is an open-source multi-agent orchestration extension developed by skyfallsin for the pi terminal coding agent ecosystem. It provides…
Agent Terminal — Agent Terminal is an open-source terminal automation harness created by Jason Kneen designed to let AI agents interact directly with…
SWE-agent — SWE-agent is an open-source autonomous software engineering agent harness built and maintained by researchers from Princeton University and Stanford University.…
Claude Squad — Claude Squad is an open-source terminal user interface (TUI) orchestrator and multi-instance session harness for Claude coding agents, developed by…
Crush — Crush is an open-source terminal coding agent created by Charm (Charmbracelet), the team behind popular Go terminal UI libraries like…