create-corral@0.4.1
latest
You don't add auth to your app.
Your agent does.
Paste one prompt. Your AI coding agent reads the spec, scaffolds auth, wires Stripe billing, adds feature gating, builds the admin dashboard. You never touch auth code.
Works with Claude Code · Cursor · Codex · OpenClaw · Windsurf · any agent that runs CLI commands
↓ Copy. Paste into your agent. Done. ↓
Read: https://llama-farm.github.io/corral/llms.txt Add auth and Stripe billing to my app. Free tier + Pro at $29/mo with 14-day trial. Wire up everything: login, signup, profile page, admin dashboard, upgrade banners, and feature gating.
Read: https://llama-farm.github.io/corral/llms.txt Add authentication to my app with email + Google login.
Read: https://llama-farm.github.io/corral/llms.txt Add Corral auth + billing to this project. Detect my framework and embed into my existing server.
The Problem
Your agent builds the UI in 30 seconds. The API in two minutes. The database schema before you finish your coffee. Then you say "add auth and payments" and watch it hallucinate for an hour. Wrong imports. Phantom endpoints. A login page that doesn't connect to anything. Every. Single. Time.
Auth-to-billing-to-gating is the hardest wiring in any app — too many moving parts, too many frameworks, too many ways to get it wrong. Your agent wasn't trained on your exact stack. It guesses. It breaks.
Corral doesn't guess. It gives your agent a spec it can actually read — every endpoint, every command, every option. One YAML config. One CLI. Login, Stripe billing, feature gating, admin dashboard, usage meters — configured in minutes, not coded over days.
You don't read auth docs. You don't write auth code. You paste a prompt, and it's done.
Why Corral
Everything your app needs to charge money and gate features. Your agent reads the spec, runs the CLI, wires the UI. You ship.
Your agent reads llms.txt, understands every endpoint, and adds auth
in 9 CLI commands. No docs to read. No Stack Overflow. Just tell it what you want.
Every command outputs --json. Built for OpenClaw, Claude Code, Codex, and Cursor.
corral.yaml defines auth providers, plans, features, and meters.
Your agent edits it, everything updates. Add a plan, add a feature gate, change
Stripe pricing — one file, zero wiring. Humans can read it too.
Traditional auth setup: 2–3 days of wiring. With Corral + your agent: one conversation. Login, billing, gating, admin dashboard — all generated. Ship the actual product, not auth plumbing.
Mounts as route handlers inside your app. Same database, same process, same deploy. No Docker. No separate service. No per-MAU pricing surprises.
Run corral stripe push and your YAML plans become Stripe
products + prices. Webhooks, checkout sessions, billing portal —
already handled.
<FeatureGate>, <PlanGate>,
<PageGate>, useFeatureGate(),
server-side gating, skeleton/blur/block modes. Every pattern you'll ever need.
Agent Workflow
Six steps from "add auth" to shipping. No human touches auth code.
llm.txt
You paste a full URL like https://corral.llamafarm.dev/llm.txt into your agent context,
or add that exact URL to AGENTS.md / .cursorrules. The agent now knows
every command, every endpoint, every option.
npm create corral
Scaffolds corral.yaml, route handlers, components, and CORRAL.md
into your project. Non-interactive. Idempotent.
corral.yaml for your plans
Free tier, Pro at $29/mo, Enterprise — whatever you asked for. Auth providers,
feature flags, usage meters. All in one file.
corral add feature, corral add provider
Incremental commands for each capability. Idempotent — safe to re-run.
--json output on every command so the agent can branch on results.
corral validate — all green
Checks config, env vars, database, Stripe sync. Exit 0 means deploy-ready.
If something's wrong, the JSON error tells the agent exactly how to fix it.
Compatibility
If it can run CLI commands and read markdown, it can use Corral.
llms.txt. It knows what to do. Add the URL to your AGENTS.md for persistent context.npm create corral + --json flags = perfect for Codex. Structured output, no screen-scraping.llms.txt URL to your .cursorrules. Cursor handles the rest from inside your editor.llms.txt to your project rules. Windsurf reads the spec and wires everything up.--json flag is on every command.One Config File
One file describes your entire auth + billing topology. Corral generates the routes, components, and database schema. Your agent edits the YAML; Corral handles the wiring.
app:
name: MySaaSApp
url: https://myapp.com
auth:
providers:
- google
- email
plans:
free:
price: 0
features: [basic_access]
pro:
price: 29
features: [*, ai_features]
trial_days: 14
features:
ai_features:
plans: [pro]
gate: blur/admin
GET /llms.txt for agents
CORRAL.md project guide
Agent-Native Design
An agent can add full auth + billing to any app in 9 commands. No screen-scraping. No ambiguity. Just structured I/O.
/api/llms.txt
Auto-generated endpoint on every Corral app. Describes the project's auth config, active plans, feature gates, and API surface in a format LLMs parse cleanly. Follows the llms.txt spec. Also served at llms.txt.
CORRAL.md
Generated at init time and updated on every corral add.
A project-specific guide: what's configured, what routes exist,
what gates are active, how to test auth locally. Designed for agent system prompts.
corral add <thing>
Incremental commands that do one thing well.
corral add feature ai_analysis --plans pro --gate blur.
Idempotent. Safe to run twice. Agent-friendly JSON output.
--json everywhere
Every CLI command accepts --json and exits with structured
output + a numeric exit code. Parse it, branch on it, pass it to the
next command. No screen-scraping.
The 9-command workflow — from zero to auth + billing:
Compatibility
Frontend frameworks, backend validators, databases, and deployment targets — Corral meets your stack.