RunWield Documentation

RunWield helps you review what an AI plans to do before it changes your code, then verifies the result. Use this manual to install RunWield, start a Session, understand its workflows, and configure it for your project.

RunWield builds on Pi. These pages explain RunWield-specific behavior and link to Pi when the behavior is unchanged. Visit the RunWield website for the product overview.

Start

Install RunWield

On macOS or Linux:

curl -fsSL https://raw.githubusercontent.com/gandazgul/runwield/main/install.sh | bash

Make sure ~/.local/bin is on your PATH, then start RunWield from your project root:

wld

On first use, connect a model provider when prompted. Then initialize the project and make a request:

/init
fix the failing parser test

• Quickstart — install, authenticate, initialize a project, and run your first request.
• Workspace — use RunWield in a browser or on your phone.

Use RunWield

• Using RunWield — commands, routing, Agents, Plans, and TUI behavior.
• Plans and workflows — triage, review, execution, validation, and recovery.
• Sessions — resume work and manage Session history.
• Self-hosted collaboration — share Plans through a Shared Space Plan Server.
• Workspace containers — run Workspace in its supported container setup.

Configure RunWield

• Providers and models — credentials, model selection, and custom providers.
• Settings reference — global and project settings.
• Customization — Agent overrides, prompts, Skills, and themes.
• Themes — select and create terminal themes.
• MCP — connect Model Context Protocol servers.

Get help

• Troubleshooting — resolve common installation and runtime problems.
• Plan lifecycle — understand durable Plan states and recovery.
• Validation authority — understand completion and review evidence.
• Contributing — build RunWield and contribute changes.

For terminal editing, keybindings, model providers, and other inherited behavior, use the Pi documentation.