openrig

Product Introduction

1. Definition: OpenRig is a multi-agent orchestration harness and coordination framework designed for long-term AI-driven software development. It is a command-line tool and TUI (Terminal User Interface) that manages a persistent team of specialized AI agents, such as Claude Code, Codex, and Pi, treating them as durable, addressable team members.
2. Core Value Proposition: OpenRig exists to solve the problem of scaling and sustaining complex coding projects over weeks or months by enabling AI agent collaboration and persistent task delegation. It moves beyond single-session AI coding assistants by creating a self-hosted AI software factory where work is owned, tracked, and coordinated between multiple specialized agents.

Main Features

1. Persistent, Addressable Agent Team: OpenRig agents are not ephemeral chat sessions. They are set up as long-running, named entities (e.g., dev-ui@factory, dev-app@factory) with their own state and communication queues. This allows for durable AI specialists that retain context and responsibility for their domain across the entire project lifecycle.
2. Structured Task Delegation & Ownership: The core orchestration mechanism uses a rig queue system. A lead agent can create a named assignment (e.g., feature-001) for a specialist agent. The specialist reads, claims, and owns that task, making its progress visible and trackable. This enforces AI work tracking and prevents context loss.
3. Direct Inter-Agent Communication (rig send): Agents can communicate directly without user intermediation for knowledge sharing. A frontend agent can query a backend specialist about an API contract using a direct message, leveraging retained project decisions without creating a new tracked task. This enables efficient AI team coordination.
4. Hierarchical Agent Spawning: Specialists can spawn temporary, bounded sub-agents (e.g., for code inspection or test generation) using native AI provider tools (like Claude's Agent feature). These sub-agents handle granular work but do not have persistent OpenRig addresses, maintaining a clean multi-agent architecture.
5. Centralized TUI Dashboard: The OpenRig TUI provides a real-time overview of the entire agent team. Users can monitor agent organization, active work queues, assignment status, and terminal outputs from a single interface, offering unified AI project management.

Problems Solved

1. Pain Point: Context Collapse in Long-Running AI Projects. Traditional AI coding assistants lose context between sessions, making it impossible to manage projects that span weeks. OpenRig solves this by giving AI agents persistent identity and memory of their assignments.
2. Pain Point: Lack of Specialization and Coordination. Using a single AI for all tasks (frontend, backend, architecture) leads to suboptimal results. OpenRig enables the creation of a specialized AI developer team where a backend agent and a UI agent can collaborate directly.
3. Target Audience: Senior Developers & Tech Leads managing large-scale refactors or greenfield projects; Small Startup Teams needing to amplify engineering capacity; Research & Development Teams prototyping complex systems with AI.
4. Use Cases: Building a full-stack application over months by delegating modules to different agents; Maintaining and extending a legacy codebase by using a dedicated agent to own its complexity; Coordinating a multi-phase project where design, implementation, and testing are handled by different AI specialists.

Unique Advantages

1. Differentiation: Unlike monolithic AI coding platforms (GitHub Copilot, Cursor) or single-session agent frameworks (AutoGPT, Smol Agents), OpenRig focuses on cross-harness, persistent coordination. It doesn't replace Claude Code or Codex; it orchestrates them as independent, collaborating entities over time, akin to a project manager for AI agents.
2. Key Innovation: The "software factory" model implemented via a simple but powerful primitives: named agents, a message queue for assignments, and direct messaging. This lightweight coordination layer turns multiple AI harnesses into a coherent, accountable team. Its self-hosted, CLI-first design prioritizes developer control and integration into existing workflows.

Frequently Asked Questions (FAQ)

1. What is OpenRig used for? OpenRig is used for orchestrating multiple AI coding agents (like Claude Code and Codex) to collaborate on long-term, complex software projects, enabling task delegation, persistent context, and coordinated work across weeks or months.
2. How does OpenRig handle AI agent context and memory? OpenRig does not manage the AI's internal context. Instead, it provides the orchestration layer by giving agents persistent identities and a work queue system. Each agent is responsible for maintaining its own context within its native environment (e.g., Claude Code session), while OpenRig tracks what work is assigned to whom.
3. Can I use OpenRig with any AI model? OpenRig primarily orchestrates AI agent harnesses, not raw models. It is designed to work with agents like Claude Code (through its CLI/API) and Codex, which have capabilities for spawning sub-agents and running code. It is agnostic to the underlying LLM as long as it runs within a compatible harness.
4. Is OpenRig a hosted service or self-hosted? OpenRig is a self-hosted, open-source CLI tool (Apache 2.0 license). You install it via npm (npm install -g @openrig/cli) and run it on your own infrastructure, giving you full control over your AI agents, data, and project security.
5. What are the main commands in the OpenRig CLI? Key commands include rig setup (configures agents), rig up (starts a project/team), rig tui (launches the dashboard), rig queue (manages task assignments), and rig send (for direct inter-agent communication).