Build, script,
and ship.
Guides for making Vibes apps, wiring up backends, and deploying with the CLI. Every app you see on this site started from a prompt — these guides show you how to make one that's yours.
Start here 3
What Is a Vibe?
A new unit of web software: where the webpage was a document and the SPA was an application, a vibe is a conjured, personal, shareable app — summoned by describing it, local by default, alive with AI, and composed from living starters and parts.
Read the explainer →How to Make & Share an App
Step-by-step screenshot walkthrough — from prompt to live, shareable app in minutes. No coding required.
Read the guide →The Platform Handles the Rest
Generated apps write only their distinctive features — identity, access, sandboxing, data, and serving are platform substrate. Why that removes the prototype wall instead of making you climb it.
Read the principle →Build 12
Your App Has a Backend
Ship a backend.js next to your app: webhook endpoints, write reactions, scheduled jobs, per-vibe secrets, outbound fetch, and server-side AI via ctx.callAI.
Read the guide →Works With
GitHub, Threads, Instagram, LinkedIn, Stripe, Zapier, Claude and ChatGPT, Muse, calendars, YouTube and Spotify. One page per service: what you can build, how it connects, and starter prompts that build in one pass.
See what it works with →Connect Accounts
One button links a visitor's own GitHub, Threads, Instagram or LinkedIn account to your app. The platform holds the credential; your backend reads through a typed library and branches on a small set of named refusals.
Connect an account →Take Payments
Let a table pay its check, take tips, charge for a booking, sell tickets, collect dues, or run a paid membership. Your Stripe account, Stripe's checkout page, and a record in your app that only your backend can mark paid.
Take payments →Patterns Cookbook
Production patterns from the vibes that run this platform: write-only credential vaults, doc-status state machines, CI lanes, self-rotating tokens, operator consoles, deploy buttons, notifications. Each one remixable.
Browse the patterns →Connect Backend Data
Read and write your app's live database from any JS backend — Node, Deno, Bun, or Cloudflare Workers.
Read the guide →Choosing AI Models
List the text and image models your app can use, let people pick one, or line up every image model on the same prompt. Your app's own choice comes first.
Pick a model →Local-first data & sync
Writes succeed locally and sync in the background. Learn what offline, revocation, blocking, and conflicts mean for your app's data.
Understand data & sync →Edit Text in Place
Click a heading, button label or sentence on your live app, type the new words, and save. Your app keeps everything else exactly as it was.
Read the guide →Connect Claude or ChatGPT
Connect your assistant to your apps, with a copyable setup prompt for hosted OAuth or the local CLI.
Read the guide →World Maps
Show travel memories, country comparisons and places to visit on a world map.
Read the guide →Your Own Style
Name the look you want in your prompt and the app is built in it, on colours you can keep adjusting.
Read the guide →CLI 3
CLI: npx vibes-diy generate
One command deploys a live React app from a prompt. 85 apps built this way. Read the workflow.
See the CLI →The CLI Handbook
Ship apps from the shell, script your data, tail changes live, manage secrets and versions, and mount your vibes into any AI agent over MCP.
Read the guide →CLI Command Reference
Every vibes-diy command, subcommand, and flag — generated from the CLI's own help output.
Browse the commands →How it works 3
How Your App Is Served
The path from URL to pixels: server-rendered first paint with your public data already in it, the edge cache and its freshness rules, deploy pins, and why private apps are never cached.
See the serve path →The System Prompts
The generator's prompts are open source on npm. What's in @vibes.diy/prompts, how they're assembled per app, and how to point your own model at them.
Read the prompts →How a Build Flows
One session, one build flow: the root owns the conversation, chooses between the builder and direct edits, delegates research, always follows a first draft with real access rules, and wraps up when everything has landed. The whole path, with the code behind each step.
Follow the flow →Reference 3
The Generator's Reference
The exact API docs the code generator reads — backend.js, Fireproof, callAI, ImgGen, and every creative skill — rendered fresh from platform source on every deploy.
Browse the reference →Access API
Private channels, anonymous submissions, role-gated reads. One file controls who can write what and who can see it — server-enforced, no backend required.
Read the API reference →Access Function Spec
The complete visual specification: the full API, worked use cases (onboarding, surveys, invite-only variants), and the per-database runtime architecture.
Open the spec →Your account 2
Credits & Top-Ups
Monthly vs purchased buckets, member-only top-ups, auto-refill, and the 'used N credits in the last hour' note that appears when your burn rate is high.
Understand credits →Notifications
Your inbox, the notification you opened, and which bells ring — including why only an explicit publish announces a release.
Read about notifications →