Documentation

Welcome. AuthRight Harness is the Identity & Access Security Decision Layer for AI software engineering: Understand → Decide → Request Match → Handoff. It helps you and your coding agents lock identity decisions, then check that the locked plan still matches the request, before you hand work to coding agents via OpenSpec or Agent Plan-mode (see Implementation).

Read these first

1. Installation — install @authright/cli from npm
2. Getting started — initialize a project and run your first Auth Change
3. How commands work — terminal authright vs /auth:… chat commands

That is enough to be productive. Everything below is optional until you need it.

Pick your path

I'm brand new. Follow the three pages above. When you want the mental model, skim Overview.

I have an existing IAM codebase. After install, jump to Existing projects and authright scan—don't document the whole world first. Java and Node.js / TypeScript are the strongest packs today; Python and Go are supported and being strengthened.

I want the day-to-day loop. Workflows and Examples.

The AI drafted decisions — now what? Reviewing decisions, then OpenSpec or Agent Plan-mode for export.

I need a command lookup. Commands (chat) and CLI (terminal).

I want to understand the system. Overview, Concepts, Implementation.

Something's broken. Troubleshooting and FAQ.

The thirty-second version

1. Install     npm install -g @authright/cli@latest
2. Initialize  cd your-project && authright init
3. Understand  /auth-scan (Cursor) or /auth:scan (Claude) or authright scan
4. Pick one       /auth:explore (HITL) or /auth:change (auto-plan), then ready on that Auth Change
5. Request match  /auth:request-match <id>
6. Handoff        pick /auth:to-openspec <id> or Plan mode + /auth:to-plan <id>
7. Verify         /auth:review <id>

Steps 1–2 and most CLI lifecycle moves happen in your terminal. Slash commands and Codex skills run in your AI assistant chat. That split is explained in How commands work.