Scion Overview

Scion is an experimental multi-agent orchestration testbed designed to manage concurrent LLM-based agents running in containers across your local machine and remote clusters. It enables developers to run groups of specialized agents with isolated identities, credentials, and workspaces, allowing for a dynamic and evolving graph of parallel execution of tasks such as research, coding, auditing, and testing.

Configuration

Scion uses a flexible configuration system based on Profiles, Runtimes, and Harnesses. This allows you to define different environments (e.g., local Docker vs. remote Kubernetes) and switch between them easily.

• Global Settings: ~/.scion/settings.yaml (User-wide defaults)
• Grove Settings: .scion/settings.yaml (Project overrides)

For detailed information on configuring Scion, see the Orchestrator Settings Reference and Agent Configuration Reference. To learn about the different agent tools supported by Scion, see Supported Harnesses.

Getting Started

Scion is designed to be easy to start with.

1. Install: Follow the Installation Guide to get Scion on your machine.
2. Initialize: Run scion init in your project root to create a .scion directory.
3. Start an Agent: Use scion start <agent-name> "<task>" to launch an agent.
4. Interact: Use scion attach <agent-name> to interact with the agent’s session, or scion logs <agent-name> to view its output.
5. Resume: Use scion resume <agent-name> to restart a stopped agent, preserving its state.

Architecture

Scion follows a Manager-Worker architecture:

• scion: A host-side CLI that orchestrates the lifecycle of agents. It manages the “Grove” (the project workspace) and provides tools for template management (scion templates).
• Agents: Isolated runtime containers (e.g., Docker) running the agent software (like Gemini CLI, Claude Code, or OpenAI Codex).