Introduction

RED is a SaaS starter kit for teams building AI products. It gives you a working app with auth, organizations, billing, admin, audit logs, a Convex backend, and an agent engine that can use tools and background tasks.

The goal is not to teach you distributed systems before you can ship. The goal is to give you a working product shape, then show you where to customize it.

The Reference Product

RED includes the Brief AI Assistant. It is a small but real example of the architecture:

• Users create a Brief workspace.
• The assistant reads and edits brief sections through tools.
• The assistant can start a Markdown export that runs in the background.
• Progress appears in the UI while the chat remains usable.
• When the export finishes, the result re-enters the thread as a task_event.

That is the product pattern RED is optimized for: collaborative AI workspaces where agents do useful work, not just answer messages.

What You Customize

Most teams start by changing three things:

1. Domain model — replace the demo entities with your own: projects, reports, invoices, deals, tickets, studies, or whatever your product manages.
2. Agents and tools — define what your assistant knows how to do in that domain.
3. Task nodes — move long work, exports, imports, or remote API calls into background tasks.

The /red-init setup skill handles the first round of project configuration. After that, you mostly work inside feature modules: backend schema/functions, app routes, agents, tools, and task nodes.

Product Foundations

RED includes the parts most AI products need before they can become a business:

• BetterAuth for email/password, OAuth, passkeys, 2FA, sessions, organizations, and invitations.
• Convex for the database, functions, reactive UI updates, scheduling, and file storage.
• Autumn and Stripe for plans, entitlements, checkout, and metered usage.
• Resend for transactional email.
• A global admin portal for users, organizations, billing, audit, waitlist, and system configuration.
• API keys for machine access when enabled.
• Audit logging and rate limiting around sensitive flows.

Stack

Layer	Technology
Frontend	React 19, Vite, Tailwind 4, shadcn/ui
Backend	Convex
Auth	BetterAuth
Billing	Autumn + Stripe
AI	AI SDK v6 with gateway routing
Email	Resend + React Email
Package manager	Bun
Monorepo	Turborepo
Linting	Biome

RED uses Convex Cloud for the backend. The frontend can be deployed anywhere that can host a Vite app.

Architecture

red-boilerplate/
├── apps/
│   ├── web/              # React frontend
│   └── docs/             # Documentation site
├── packages/
│   ├── backend/          # Convex backend functions and schema
│   ├── backend-contract/ # Frontend-safe backend constants and generated API
│   ├── web-core/         # Shared hooks, auth client, and utilities
│   ├── web-ui/           # shadcn/ui component library
│   ├── web-shell/        # App shell and shared layout
│   └── web-admin/        # Global admin portal

The frontend imports backend types and API references through @red/backend/api and @red/backend/dataModel. Shared frontend packages use @red/web-core, @red/web-ui, @red/web-shell, and @red/web-admin.

Organization Isolation

Everything in RED is scoped to organizations. Each organization has its own:

• Members and roles.
• Billing subscription and usage state.
• AI threads, boundaries, and conversation history.
• Audit trail.
• API keys.

The global admin can control self-signup, waitlist mode, and whether users can create organizations from the Configure panel.

Where To Go Next

• Quick Start gets the app running.
• Initialize Your Project explains the /red-init skill.
• Core Mental Model explains agents, tools, task nodes, and events.
• Building AI Features gives source-backed recipes for extending the AI module.

FAQ

Can I build a non-AI SaaS with RED? Yes. The AI module can be removed during initialization. Auth, organizations, admin, audit, email, and billing still make RED useful for normal SaaS products.

Can I add a marketing site? Yes. The apps/ directory is a Turborepo workspace, so you can add another app alongside apps/web.

Does RED include RAG? Not out of the box. RED is agentic-first: agents use explicit tools to read and act on data. If you need vector retrieval, add it as a tool or inject retrieved context into the agent instructions.

How do I configure pricing and plans? Configure pricing in Autumn, then sync the billing state from RED’s admin area.